-
Notifications
You must be signed in to change notification settings - Fork 95
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merge NN (tf2.1) #330
Merge NN (tf2.1) #330
Commits on Feb 5, 2020
-
Add instance cropping transformer.
- Agnostic to centroid calculation method. - Propagates all frame example metadata to individual instance examples. - Tests for coordinate transformations correctness, especially for tricky bounding box and cropping calculations.
Configuration menu - View commit details
-
Copy full SHA for a781060 - Browse repository at this point
Copy the full SHA a781060View commit details
Commits on Feb 6, 2020
-
Add confidence map generation transformers.
- Low level confidence map generation functions - Multi-instance confidence map generator - Multi-instance centroid confidence map generator - Instance-cropped center instance confidence map generator - Instance-cropped all instance confidence map generator - Tests
Configuration menu - View commit details
-
Copy full SHA for 7579d23 - Browse repository at this point
Copy the full SHA 7579d23View commit details -
Store raw image size in examples generated by data providers.
- Useful for downstream processing that affects shape.
Configuration menu - View commit details
-
Copy full SHA for 14a8230 - Browse repository at this point
Copy the full SHA 14a8230View commit details -
Rename instance cropping image size to "full_image_*"
- "Full" to indicate it refers to the input to the instance cropper, which may have been modified relative to the "raw" image, e.g., due to padding/scaling. - Rename attribute `drop_full_image` to `keep_full_image`
Configuration menu - View commit details
-
Copy full SHA for 7763cdd - Browse repository at this point
Copy the full SHA 7763cddView commit details -
Modified the "scale" key to a 2-tuple indicating the x- and y-scaling.
- The general expectation is that points in individual examples are already on the image grid scale. - The "scale" key can be used to track scaling operations, including aspect ratio changes. This information is sufficient to map points on example images/confidence maps to the raw images. - Commented scaling behavior across pipeline modules. - Changed behavior of `MultiConfidenceMaps` to compute confidence maps on the current image grid under the assumption that the "instances" key is at the same scale as the "image".
Configuration menu - View commit details
-
Copy full SHA for 0499e98 - Browse repository at this point
Copy the full SHA 0499e98View commit details -
Initial data normalization method implementations.
- Documented the normalization modes expected when using different pretrained models from tf.keras.applications. - Still needs tests and a transformer block class.
Configuration menu - View commit details
-
Copy full SHA for e705862 - Browse repository at this point
Copy the full SHA e705862View commit details -
Configuration menu - View commit details
-
Copy full SHA for 3e2c68a - Browse repository at this point
Copy the full SHA 3e2c68aView commit details -
Configuration menu - View commit details
-
Copy full SHA for 23057ed - Browse repository at this point
Copy the full SHA 23057edView commit details
Commits on Feb 7, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 2380b47 - Browse repository at this point
Copy the full SHA 2380b47View commit details -
Configuration menu - View commit details
-
Copy full SHA for cbcbde5 - Browse repository at this point
Copy the full SHA cbcbde5View commit details -
Configuration menu - View commit details
-
Copy full SHA for f8f9cb9 - Browse repository at this point
Copy the full SHA f8f9cb9View commit details
Commits on Feb 9, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 79d2632 - Browse repository at this point
Copy the full SHA 79d2632View commit details -
Configuration menu - View commit details
-
Copy full SHA for 5a49b73 - Browse repository at this point
Copy the full SHA 5a49b73View commit details -
Configuration menu - View commit details
-
Copy full SHA for 9d5cadc - Browse repository at this point
Copy the full SHA 9d5cadcView commit details -
Configuration menu - View commit details
-
Copy full SHA for 5e12702 - Browse repository at this point
Copy the full SHA 5e12702View commit details -
Configuration menu - View commit details
-
Copy full SHA for 80422dc - Browse repository at this point
Copy the full SHA 80422dcView commit details -
Configuration menu - View commit details
-
Copy full SHA for bdb6263 - Browse repository at this point
Copy the full SHA bdb6263View commit details
Commits on Feb 10, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 3f66b4c - Browse repository at this point
Copy the full SHA 3f66b4cView commit details -
Edge maps transformers module.
- Implements edge confidence maps; equivalent of part confidence maps where scalar values are scaled by the distance of the sampling grid points to the line segment formed by edges. - Implement part affinity fields as simple unit vectors weighted by the edge confidence maps for masking.
Configuration menu - View commit details
-
Copy full SHA for e6362d5 - Browse repository at this point
Copy the full SHA e6362d5View commit details -
Configuration menu - View commit details
-
Copy full SHA for 3106934 - Browse repository at this point
Copy the full SHA 3106934View commit details -
Configuration menu - View commit details
-
Copy full SHA for 41e3c43 - Browse repository at this point
Copy the full SHA 41e3c43View commit details -
Configuration menu - View commit details
-
Copy full SHA for 2407151 - Browse repository at this point
Copy the full SHA 2407151View commit details -
Configuration menu - View commit details
-
Copy full SHA for 8bb36b9 - Browse repository at this point
Copy the full SHA 8bb36b9View commit details -
Configuration menu - View commit details
-
Copy full SHA for 58d134a - Browse repository at this point
Copy the full SHA 58d134aView commit details -
Configuration menu - View commit details
-
Copy full SHA for 747c901 - Browse repository at this point
Copy the full SHA 747c901View commit details -
Configuration menu - View commit details
-
Copy full SHA for a88ae41 - Browse repository at this point
Copy the full SHA a88ae41View commit details -
Configuration menu - View commit details
-
Copy full SHA for a296f21 - Browse repository at this point
Copy the full SHA a296f21View commit details -
Configuration menu - View commit details
-
Copy full SHA for 91a38b5 - Browse repository at this point
Copy the full SHA 91a38b5View commit details
Commits on Feb 12, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 278b15d - Browse repository at this point
Copy the full SHA 278b15dView commit details -
Dataset operations transformers module
- Transformers that operate over multiple elements of the data: - Shuffler - Batcher (this one is tricky with variable length elements) - Prefetcher
Configuration menu - View commit details
-
Copy full SHA for f624e57 - Browse repository at this point
Copy the full SHA f624e57View commit details -
Add initial generic pipelines for generating training data.
- BottomUpPipeline: for building standard multi-confmap + PAF pipelines - TopDownPipeline: for building standard instance cropped pipelines
Configuration menu - View commit details
-
Copy full SHA for 6ba0a62 - Browse repository at this point
Copy the full SHA 6ba0a62View commit details -
Fix truncation rounding error during instance cropping.
- float32 precision sometimes results in bounding box sizes that are #.9997, which can result in an incorrect crop, even though the centered bounding boxes are created with exact dimensions.
Configuration menu - View commit details
-
Copy full SHA for b71be9f - Browse repository at this point
Copy the full SHA b71be9fView commit details -
Configuration menu - View commit details
-
Copy full SHA for a9662d7 - Browse repository at this point
Copy the full SHA a9662d7View commit details -
Configuration menu - View commit details
-
Copy full SHA for d3784cc - Browse repository at this point
Copy the full SHA d3784ccView commit details -
Configuration menu - View commit details
-
Copy full SHA for eaff014 - Browse repository at this point
Copy the full SHA eaff014View commit details -
Configuration menu - View commit details
-
Copy full SHA for 521559c - Browse repository at this point
Copy the full SHA 521559cView commit details -
- Loads elements of the pipeline at creation time rather than during execution. This is a workaround to weird thread safety issues with using Labels as a provider during training.
Configuration menu - View commit details
-
Copy full SHA for c120486 - Browse repository at this point
Copy the full SHA c120486View commit details
Commits on Feb 13, 2020
-
Fix stochastic deadlocks in tf.py_functions.
- These resulted random hangs during training that could not be interrupted via SIGINT or otherwise. - Solution was to remove TF ops from inside py_functions in the data pipeline. - Interesting side effect is a big performance speedup. - See: tensorflow/tensorflow#32454
Configuration menu - View commit details
-
Copy full SHA for 613c201 - Browse repository at this point
Copy the full SHA 613c201View commit details -
Configuration menu - View commit details
-
Copy full SHA for 6ccf5d2 - Browse repository at this point
Copy the full SHA 6ccf5d2View commit details -
Configuration menu - View commit details
-
Copy full SHA for 5a4ea16 - Browse repository at this point
Copy the full SHA 5a4ea16View commit details -
Configuration menu - View commit details
-
Copy full SHA for 9ba4d25 - Browse repository at this point
Copy the full SHA 9ba4d25View commit details -
Configuration menu - View commit details
-
Copy full SHA for 83891cf - Browse repository at this point
Copy the full SHA 83891cfView commit details -
Added
example_indices
attribute to providers.- Replaces the `ds_index` argument to the `make_dataset` method - Allows for flexible specification of subsets of the data to be used without having to create new source data objects (e.g., sleap.Labels). This makes it easier to read subsets of videos or to split the labels for train/val/test sets. - Examples generated by `LabelsReader` now include an "example_ind" key to indicate the index of the example within the original data.
Configuration menu - View commit details
-
Copy full SHA for 561e95f - Browse repository at this point
Copy the full SHA 561e95fView commit details -
Configuration menu - View commit details
-
Copy full SHA for 74bc2c9 - Browse repository at this point
Copy the full SHA 74bc2c9View commit details -
Initial nn.config implementation.
- This contains simple attr attribute classes with all user-facing configuration options related to training. - This supersedes the previous TrainingJob class and should contain all serializable job-related parameters. - Functional modules should be able to be instantiated from these, but can be decoupled in the actual implementation.
Configuration menu - View commit details
-
Copy full SHA for 507a051 - Browse repository at this point
Copy the full SHA 507a051View commit details -
Pipeline attributes default to empty lists.
- This is necessary to build partial pipelines without providers.
Configuration menu - View commit details
-
Copy full SHA for ae1c555 - Browse repository at this point
Copy the full SHA ae1c555View commit details -
Configuration menu - View commit details
-
Copy full SHA for d91865d - Browse repository at this point
Copy the full SHA d91865dView commit details -
Configuration menu - View commit details
-
Copy full SHA for 0637835 - Browse repository at this point
Copy the full SHA 0637835View commit details -
Configuration menu - View commit details
-
Copy full SHA for 44d2cf1 - Browse repository at this point
Copy the full SHA 44d2cf1View commit details -
Configuration menu - View commit details
-
Copy full SHA for 74e0391 - Browse repository at this point
Copy the full SHA 74e0391View commit details -
Configuration menu - View commit details
-
Copy full SHA for ac59f24 - Browse repository at this point
Copy the full SHA ac59f24View commit details -
Configuration menu - View commit details
-
Copy full SHA for 53d538b - Browse repository at this point
Copy the full SHA 53d538bView commit details -
Configuration menu - View commit details
-
Copy full SHA for 0d2ec0c - Browse repository at this point
Copy the full SHA 0d2ec0cView commit details -
Configuration menu - View commit details
-
Copy full SHA for 10d8427 - Browse repository at this point
Copy the full SHA 10d8427View commit details -
Major nn refactor in preparation for new Model/ModelConfig classes
- Cleaned up nn.architectures to remove architectures that need major updating before they can be ported, removed unused utility functions. - We'll port in the deleted architectures from other branches when we're ready, but in the meanwhile they can't be used without major rewrites. - Exception: hrnet module which is not pretrained and will require minor adjustments to fit new format. - Replace old nn.model module with nn.model_config - Removed PreprocessingConfig from nn.model since that's covered elsewhere in nn.config.DataConfig. - Removed preprocessing-related args from legacy classmethod builders in ModelConfig.
Configuration menu - View commit details
-
Copy full SHA for 7323963 - Browse repository at this point
Copy the full SHA 7323963View commit details -
Add attribute for specifying UNet stem kernel size.
- Useful for increasing receptive field without adding more blocks. - Defaults still match the canonical architecture and tests.
Configuration menu - View commit details
-
Copy full SHA for 3fe10d7 - Browse repository at this point
Copy the full SHA 3fe10d7View commit details -
Split nn.config.py into submodules for legibility + initial ModelConfig
- Config sections now sit in their own submodules, but the syntax is the same since all config classes are imported in nn/config/__init__.py. - Initial implementation of ModelConfig. Totally decoupled from previous Model/ModelConfig implementations. - High level architecture parameters are now explicitly specified for each architecture and "oneof"-style options will just be specified by having different attributes which validators can check for mutual exclusivity. This should make it easier to serialize into more readable json as opposed to the old system of specifying the class name and doing some voodoo to construct the correct one. - This decoupling comes at the cost of reduced flexibility in model customization (though this can overcome with "generic" model configs like EncoderDecoder), as well as requiring manual specification of parameters for each architecture.
Configuration menu - View commit details
-
Copy full SHA for 248ecb1 - Browse repository at this point
Copy the full SHA 248ecb1View commit details -
Configuration menu - View commit details
-
Copy full SHA for ba205d8 - Browse repository at this point
Copy the full SHA ba205d8View commit details -
Implement oneof decorator to enforce mutual exclusivity
- This achieves validation to enforce the behavior desired for the mutually-exclusive attribute classes. - Classes decorated with `attr.s` and `oneof` will raise an error if more than one attribute is set to not None when instantiated.
Configuration menu - View commit details
-
Copy full SHA for 7f1f67c - Browse repository at this point
Copy the full SHA 7f1f67cView commit details
Commits on Feb 17, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 1de72cd - Browse repository at this point
Copy the full SHA 1de72cdView commit details
Commits on Feb 18, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 88ec80f - Browse repository at this point
Copy the full SHA 88ec80fView commit details -
Configuration menu - View commit details
-
Copy full SHA for 893758c - Browse repository at this point
Copy the full SHA 893758cView commit details -
Configuration menu - View commit details
-
Copy full SHA for 9845738 - Browse repository at this point
Copy the full SHA 9845738View commit details -
Configuration menu - View commit details
-
Copy full SHA for 946310a - Browse repository at this point
Copy the full SHA 946310aView commit details -
Configuration menu - View commit details
-
Copy full SHA for 70b139f - Browse repository at this point
Copy the full SHA 70b139fView commit details -
Configuration menu - View commit details
-
Copy full SHA for 160d8db - Browse repository at this point
Copy the full SHA 160d8dbView commit details -
Configuration menu - View commit details
-
Copy full SHA for 922cb6f - Browse repository at this point
Copy the full SHA 922cb6fView commit details -
Configuration menu - View commit details
-
Copy full SHA for cf3cdf3 - Browse repository at this point
Copy the full SHA cf3cdf3View commit details
Commits on Feb 19, 2020
-
Refactor nn.heads and nn.model
- classmethods for instantiation from config classes - Moved old modules to *_legacy.py until TrainingJob conversion is done - Several tweaks to configs
Configuration menu - View commit details
-
Copy full SHA for 67286bf - Browse repository at this point
Copy the full SHA 67286bfView commit details -
Configuration menu - View commit details
-
Copy full SHA for f683c46 - Browse repository at this point
Copy the full SHA f683c46View commit details -
Add global structuring hooks for sleap.Skeleton
- This allows for (un)structuring instances without needing to write any extra code for instances nested in larger attrs classes.
Configuration menu - View commit details
-
Copy full SHA for 0fc1260 - Browse repository at this point
Copy the full SHA 0fc1260View commit details -
Configuration menu - View commit details
-
Copy full SHA for 252c9ff - Browse repository at this point
Copy the full SHA 252c9ffView commit details -
Add profile_graph to TensorBoard configuration.
- Removed part_metrics since this will always be enabled.
Configuration menu - View commit details
-
Copy full SHA for da8a050 - Browse repository at this point
Copy the full SHA da8a050View commit details -
Create nn.losses module for custom losses/metrics
- Created standalone OHKM loss computation function for easier testing. - Added heuristics for ratio-based dynamic hard keypoint detection. - Fix default OHKM ratio config (should be >1)
Configuration menu - View commit details
-
Copy full SHA for 9aa3b24 - Browse repository at this point
Copy the full SHA 9aa3b24View commit details -
Created initial transformer classes for inference
- KerasModelPredictor supports multi-input/output and runs extremely fast on GPU (but the device can be set at the class-level). - Refactored global peak finding to return coordinates in points format and apply thresholding before returning. - GlobalPeakFinder transformer for generic peak finding on any confmaps key with stride adjustment.
Configuration menu - View commit details
-
Copy full SHA for 458e1f9 - Browse repository at this point
Copy the full SHA 458e1f9View commit details -
Instance crop size estimation utility function
- Should probably get moved somewhere else when nn is cleaned up
Configuration menu - View commit details
-
Copy full SHA for ec33486 - Browse repository at this point
Copy the full SHA ec33486View commit details -
Configuration menu - View commit details
-
Copy full SHA for a7902cd - Browse repository at this point
Copy the full SHA a7902cdView commit details -
Configuration menu - View commit details
-
Copy full SHA for c579480 - Browse repository at this point
Copy the full SHA c579480View commit details -
Initial nn.training module refactoring
- Moved old training module to *_legacy.py for now - Added several generic builder functions. These will likely be replaced by the TrainingJob/Trainer classes, but writing them out now to get a sense of where we need coupling.
Configuration menu - View commit details
-
Copy full SHA for dd5dcde - Browse repository at this point
Copy the full SHA dd5dcdeView commit details -
Configuration menu - View commit details
-
Copy full SHA for 75d8718 - Browse repository at this point
Copy the full SHA 75d8718View commit details -
Configuration menu - View commit details
-
Copy full SHA for c3d7071 - Browse repository at this point
Copy the full SHA c3d7071View commit details -
Configuration menu - View commit details
-
Copy full SHA for 3676206 - Browse repository at this point
Copy the full SHA 3676206View commit details -
Configuration menu - View commit details
-
Copy full SHA for aa03885 - Browse repository at this point
Copy the full SHA aa03885View commit details -
- Previously the first stem block started with a max pooling directly on the input layer
Configuration menu - View commit details
-
Copy full SHA for 1fcd217 - Browse repository at this point
Copy the full SHA 1fcd217View commit details -
Configuration menu - View commit details
-
Copy full SHA for 5cd954c - Browse repository at this point
Copy the full SHA 5cd954cView commit details -
Configuration menu - View commit details
-
Copy full SHA for 77c5345 - Browse repository at this point
Copy the full SHA 77c5345View commit details -
Bug fixes in KeyMapper and enforcement of static shapes
- Should probably look into this issue more closely since it breaks dynamic shaped inputs (could be an issue for multi-instance inference potentially?)
Configuration menu - View commit details
-
Copy full SHA for 1800f27 - Browse repository at this point
Copy the full SHA 1800f27View commit details -
Implement data pipelines for the 3 major model types
- Implements several versions of each pipeline: - base_pipeline: Fetches input data with no optimization. Useful for composing with other pipelines and for quick inspection. - training_pipeline: Generates examples with all data structures needed for training. Uses optimizations to minimize I/O latency across iterations like preloading/prefetching. Applies data augmentation and shuffling. - viz_pipeline: Produces minimal examples necessary for generating visualizations of predictions during training. Necessarily, this implements basic inference, but does not have some optimizations that a full inference pipeline might benefit from. - Lots of repetition but keeping it as-is for now for flexibility
Configuration menu - View commit details
-
Copy full SHA for 33055cc - Browse repository at this point
Copy the full SHA 33055ccView commit details -
Configuration menu - View commit details
-
Copy full SHA for 4e46634 - Browse repository at this point
Copy the full SHA 4e46634View commit details -
Configuration menu - View commit details
-
Copy full SHA for 00bbe94 - Browse repository at this point
Copy the full SHA 00bbe94View commit details
Commits on Feb 20, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 8d56729 - Browse repository at this point
Copy the full SHA 8d56729View commit details -
Configuration menu - View commit details
-
Copy full SHA for b0996d1 - Browse repository at this point
Copy the full SHA b0996d1View commit details
Commits on Feb 21, 2020
-
Add OptionalSpinWidget for optional int/doubles.
If field type is optional_int (optional_double), then we'll show a widget with a spin box (for int or double) and a checkbox to disable the spin box. Box is checked iff field value is None.
Configuration menu - View commit details
-
Copy full SHA for 7a6e8b0 - Browse repository at this point
Copy the full SHA 7a6e8b0View commit details -
Configuration menu - View commit details
-
Copy full SHA for d798457 - Browse repository at this point
Copy the full SHA d798457View commit details -
Add TextOrListWidget for text/list widget.
If options are set then will show as a list (i.e., QComboBox), otherwise will be a text field (i.e., QLineEdit).
Configuration menu - View commit details
-
Copy full SHA for 5e24385 - Browse repository at this point
Copy the full SHA 5e24385View commit details
Commits on Feb 24, 2020
-
- support for "text" field type - for optional field types, return specified string for val when None - add set_field_enabled method - emit valueChanged when fields in StackBuilderWidget changed
Configuration menu - View commit details
-
Copy full SHA for 44b1c45 - Browse repository at this point
Copy the full SHA 44b1c45View commit details -
Configuration menu - View commit details
-
Copy full SHA for 8502a24 - Browse repository at this point
Copy the full SHA 8502a24View commit details -
Configuration menu - View commit details
-
Copy full SHA for b5a006a - Browse repository at this point
Copy the full SHA b5a006aView commit details -
Configuration menu - View commit details
-
Copy full SHA for 4a31425 - Browse repository at this point
Copy the full SHA 4a31425View commit details -
Configuration menu - View commit details
-
Copy full SHA for 7aee059 - Browse repository at this point
Copy the full SHA 7aee059View commit details -
Configuration menu - View commit details
-
Copy full SHA for 94d1780 - Browse repository at this point
Copy the full SHA 94d1780View commit details -
Configuration menu - View commit details
-
Copy full SHA for 9c65ff7 - Browse repository at this point
Copy the full SHA 9c65ff7View commit details
Commits on Feb 25, 2020
-
- Squeeze singleton image channel (grayscale) - Explicit control of paired peak line plotting - Formatting
Configuration menu - View commit details
-
Copy full SHA for 3ddc173 - Browse repository at this point
Copy the full SHA 3ddc173View commit details -
Add docstrings to nn.model and expose keras model as attribute.
- Now Model.make_model(...) also sets that instance's keras_model attribute for convenience. - Class method constructor from config can optionally update the input config object with argument or inferred data.
Configuration menu - View commit details
-
Copy full SHA for 4cc7949 - Browse repository at this point
Copy the full SHA 4cc7949View commit details -
Add serialization/deserialization helper methods to TrainingJobConfig.
- Need to re-implement the legacy format parser when unstructuring.
Configuration menu - View commit details
-
Copy full SHA for 67884c1 - Browse repository at this point
Copy the full SHA 67884c1View commit details -
nn.training.Trainer implementation.
- Handles centroids, topdown and multi-instance model types. - Output type specific trainers: CentroidConfmapsModelTrainer, TopdownConfmapsModelTrainer, BottomUpModelTrainer - CLI with behavior mostly equivalent to old format. Run path should be fully specified since it'll be harder to parse from stdout. - Basic docstrings - Several small functions for setting up the training still need to be rolled into Trainer, or maybe a TrainingJob that sits between the config and the actual trainer.
Configuration menu - View commit details
-
Copy full SHA for 77c3cbb - Browse repository at this point
Copy the full SHA 77c3cbbView commit details -
Configuration menu - View commit details
-
Copy full SHA for 0809203 - Browse repository at this point
Copy the full SHA 0809203View commit details -
Remove unnecessary matplotlib backend specification.
- Will probably need to check that this still works across platforms, or provide a parameter to set this explicitly.
Configuration menu - View commit details
-
Copy full SHA for 35a0ae7 - Browse repository at this point
Copy the full SHA 35a0ae7View commit details -
Configuration menu - View commit details
-
Copy full SHA for 2aa2c4f - Browse repository at this point
Copy the full SHA 2aa2c4fView commit details -
Configuration menu - View commit details
-
Copy full SHA for 29cf1b8 - Browse repository at this point
Copy the full SHA 29cf1b8View commit details -
Configuration menu - View commit details
-
Copy full SHA for 7c6a853 - Browse repository at this point
Copy the full SHA 7c6a853View commit details -
Added utility functions for querying and setting GPU availability
- Check for GPUs available - Set which GPUs to use (or CPU only) - Set memory preallocation policy - Check for initialization - Explicit device initialization
Configuration menu - View commit details
-
Copy full SHA for ec596e4 - Browse repository at this point
Copy the full SHA ec596e4View commit details -
Configuration menu - View commit details
-
Copy full SHA for 8067112 - Browse repository at this point
Copy the full SHA 8067112View commit details -
Configuration menu - View commit details
-
Copy full SHA for 3e041ef - Browse repository at this point
Copy the full SHA 3e041efView commit details -
Configuration menu - View commit details
-
Copy full SHA for b6d07a1 - Browse repository at this point
Copy the full SHA b6d07a1View commit details -
Remove remaining Union-style configs to enable structuring from file
- Also fix the global structuring hook for Skeleton
Configuration menu - View commit details
-
Copy full SHA for 0426e55 - Browse repository at this point
Copy the full SHA 0426e55View commit details -
Configuration menu - View commit details
-
Copy full SHA for 7861bb5 - Browse repository at this point
Copy the full SHA 7861bb5View commit details -
Configuration menu - View commit details
-
Copy full SHA for 2628701 - Browse repository at this point
Copy the full SHA 2628701View commit details -
Configuration menu - View commit details
-
Copy full SHA for d62cb10 - Browse repository at this point
Copy the full SHA d62cb10View commit details -
Configuration menu - View commit details
-
Copy full SHA for 6699fbc - Browse repository at this point
Copy the full SHA 6699fbcView commit details -
Configuration menu - View commit details
-
Copy full SHA for 3d0563e - Browse repository at this point
Copy the full SHA 3d0563eView commit details -
Initial inference pipeline implementation
- Topdown with centroid + confmap models - Moved old inference.py until deprecated
Configuration menu - View commit details
-
Copy full SHA for a2e4c20 - Browse repository at this point
Copy the full SHA a2e4c20View commit details -
Configuration menu - View commit details
-
Copy full SHA for 43c8836 - Browse repository at this point
Copy the full SHA 43c8836View commit details -
Add more logging to Trainer and disable PAF quiver plot
- Quiver plot has a huge memory leak on the matplotlib side right now
Configuration menu - View commit details
-
Copy full SHA for 86ad90d - Browse repository at this point
Copy the full SHA 86ad90dView commit details -
Configuration menu - View commit details
-
Copy full SHA for 201b7a1 - Browse repository at this point
Copy the full SHA 201b7a1View commit details -
Configuration menu - View commit details
-
Copy full SHA for b6e3a31 - Browse repository at this point
Copy the full SHA b6e3a31View commit details -
Configuration menu - View commit details
-
Copy full SHA for 8bb7ecf - Browse repository at this point
Copy the full SHA 8bb7ecfView commit details -
Configuration menu - View commit details
-
Copy full SHA for be20671 - Browse repository at this point
Copy the full SHA be20671View commit details -
Configuration menu - View commit details
-
Copy full SHA for 0d05976 - Browse repository at this point
Copy the full SHA 0d05976View commit details -
Configuration menu - View commit details
-
Copy full SHA for 0dea1e2 - Browse repository at this point
Copy the full SHA 0dea1e2View commit details
Commits on Feb 26, 2020
-
Configuration menu - View commit details
-
Copy full SHA for a67da57 - Browse repository at this point
Copy the full SHA a67da57View commit details -
Configuration menu - View commit details
-
Copy full SHA for bb5b7c2 - Browse repository at this point
Copy the full SHA bb5b7c2View commit details -
Add best_logical_device_name utility
- Returns the best initialized (logical) device name for use with tf.device for explicit tensor device placement.
Configuration menu - View commit details
-
Copy full SHA for f620306 - Browse repository at this point
Copy the full SHA f620306View commit details -
Configuration menu - View commit details
-
Copy full SHA for be50f17 - Browse repository at this point
Copy the full SHA be50f17View commit details -
Rewrote PAF grouping module to for data pipeline inference
- PartAffinityFieldInstanceGrouper is the transformer - Pretty under-commented and incomplete typing - PAFScorer should definitely be broken up somehow
Configuration menu - View commit details
-
Copy full SHA for 7dfe7fc - Browse repository at this point
Copy the full SHA 7dfe7fcView commit details -
Configuration menu - View commit details
-
Copy full SHA for e71dbeb - Browse repository at this point
Copy the full SHA e71dbebView commit details -
Configuration menu - View commit details
-
Copy full SHA for c1f023c - Browse repository at this point
Copy the full SHA c1f023cView commit details -
Configuration menu - View commit details
-
Copy full SHA for 8914b2e - Browse repository at this point
Copy the full SHA 8914b2eView commit details -
Configuration menu - View commit details
-
Copy full SHA for 3f9d8bf - Browse repository at this point
Copy the full SHA 3f9d8bfView commit details -
Expand transformers for topdown postprocessing
- PredictedInstanceCropper and PredictedCenterInstanceNormalizer now pass forward centroid and peak confidences.
Configuration menu - View commit details
-
Copy full SHA for d7d5a14 - Browse repository at this point
Copy the full SHA d7d5a14View commit details -
Full inference pipelines implemented for topdown and bottomup.
- Both can be initialized from just run paths. - Both have a .predict() method that creates LabeledFrames on demand. - Both have a .predict_generator() method that generates raw examples for flexibility in postprocessing.
Configuration menu - View commit details
-
Copy full SHA for 6ea8d6f - Browse repository at this point
Copy the full SHA 6ea8d6fView commit details -
- Remove all unused or orphaned modules or functions. - Add basic module-level docstrings. - Minor formatting - Import most submodules from top level sleap __init__ (adds only ~400ms startup time to "import sleap")
Configuration menu - View commit details
-
Copy full SHA for c92d802 - Browse repository at this point
Copy the full SHA c92d802View commit details -
Configuration menu - View commit details
-
Copy full SHA for 3791ce4 - Browse repository at this point
Copy the full SHA 3791ce4View commit details -
Configuration menu - View commit details
-
Copy full SHA for dd360d1 - Browse repository at this point
Copy the full SHA dd360d1View commit details -
Configuration menu - View commit details
-
Copy full SHA for 5d077ed - Browse repository at this point
Copy the full SHA 5d077edView commit details -
Configuration menu - View commit details
-
Copy full SHA for c14a8c2 - Browse repository at this point
Copy the full SHA c14a8c2View commit details -
Configuration menu - View commit details
-
Copy full SHA for 801119f - Browse repository at this point
Copy the full SHA 801119fView commit details -
Configuration menu - View commit details
-
Copy full SHA for 4105678 - Browse repository at this point
Copy the full SHA 4105678View commit details -
Configuration menu - View commit details
-
Copy full SHA for 4f9fea0 - Browse repository at this point
Copy the full SHA 4f9fea0View commit details -
Configuration menu - View commit details
-
Copy full SHA for f7712ac - Browse repository at this point
Copy the full SHA f7712acView commit details -
Configuration menu - View commit details
-
Copy full SHA for e2d27e7 - Browse repository at this point
Copy the full SHA e2d27e7View commit details -
- Add pooling block at the end of the encoder stack. This is necessary since UNet uses the last conv in each block as the skip source, so the encoder blocks cannot end with a max pooling. Additionally, the first block cannot start with a max pooling on the input, so we add an extra pooling-only block at the end of the encoder stack to compensate. - Middle block implemented. This block adds convolutions without pooling at the end of the encoder stack. It increases the number of filters, and then decreases back to the input filter count. When set to True, the model construction is exactly identical to the legacy models. - Added comment about middle block to config. - Fixed incorrect upsampling blocks: should not have batch norm and the filters after the first refinement conv should decrease (within the block). - Implemented several of the functionalities required above in the simple encoder/decoder blocks.
Configuration menu - View commit details
-
Copy full SHA for 33339cd - Browse repository at this point
Copy the full SHA 33339cdView commit details
Commits on Feb 27, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 977845e - Browse repository at this point
Copy the full SHA 977845eView commit details -
Configuration menu - View commit details
-
Copy full SHA for 9aa1c2a - Browse repository at this point
Copy the full SHA 9aa1c2aView commit details -
Configuration menu - View commit details
-
Copy full SHA for 12437f7 - Browse repository at this point
Copy the full SHA 12437f7View commit details -
Configuration menu - View commit details
-
Copy full SHA for 069b156 - Browse repository at this point
Copy the full SHA 069b156View commit details -
Configuration menu - View commit details
-
Copy full SHA for ee2d923 - Browse repository at this point
Copy the full SHA ee2d923View commit details -
Enable training dialog from app gui.
No inference (w/ saved models) dialog yet, and no loading of saved configs.
Configuration menu - View commit details
-
Copy full SHA for 4ba9a47 - Browse repository at this point
Copy the full SHA 4ba9a47View commit details
Commits on Mar 1, 2020
-
Doesn't yet indicate when you've made changes to a config (ideally when you start making changes, the menu will switch to a "user config" option or something like that), and doesn't provide feedback if the file you select doesn't give you a valid config (either because it's not a valid config or because it's for the wrong head type).
Configuration menu - View commit details
-
Copy full SHA for 37ecb86 - Browse repository at this point
Copy the full SHA 37ecb86View commit details -
Configuration menu - View commit details
-
Copy full SHA for 2b663c5 - Browse repository at this point
Copy the full SHA 2b663c5View commit details -
Configuration menu - View commit details
-
Copy full SHA for 6e44b91 - Browse repository at this point
Copy the full SHA 6e44b91View commit details
Commits on Mar 2, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 9c52510 - Browse repository at this point
Copy the full SHA 9c52510View commit details -
Configuration menu - View commit details
-
Copy full SHA for 2a6f988 - Browse repository at this point
Copy the full SHA 2a6f988View commit details -
Configuration menu - View commit details
-
Copy full SHA for 3787600 - Browse repository at this point
Copy the full SHA 3787600View commit details -
Configuration menu - View commit details
-
Copy full SHA for 651653a - Browse repository at this point
Copy the full SHA 651653aView commit details -
Configuration menu - View commit details
-
Copy full SHA for de5c57f - Browse repository at this point
Copy the full SHA de5c57fView commit details -
Configuration menu - View commit details
-
Copy full SHA for 84080fd - Browse repository at this point
Copy the full SHA 84080fdView commit details -
Add UI for "use trained model"
functionality to actually use trained model not yet implemented
Configuration menu - View commit details
-
Copy full SHA for 4b9fb4c - Browse repository at this point
Copy the full SHA 4b9fb4cView commit details
Commits on Mar 3, 2020
-
Accept path to json config in cli.
If path is to config file json, then get the path to parent dir.
Configuration menu - View commit details
-
Copy full SHA for b81efc8 - Browse repository at this point
Copy the full SHA b81efc8View commit details -
Configuration menu - View commit details
-
Copy full SHA for 758048a - Browse repository at this point
Copy the full SHA 758048aView commit details -
Configuration menu - View commit details
-
Copy full SHA for 986ede9 - Browse repository at this point
Copy the full SHA 986ede9View commit details -
Configuration menu - View commit details
-
Copy full SHA for 280294c - Browse repository at this point
Copy the full SHA 280294cView commit details -
Configuration menu - View commit details
-
Copy full SHA for ed9f26d - Browse repository at this point
Copy the full SHA ed9f26dView commit details
Commits on Mar 4, 2020
-
Don't raise error in ensure_grayscale.
If we raise error when shape isn't what we want (1 or 3), then this error is getting raised whenever function is called during transform on tf dataset. ensure_rgb() already didn't raise error for unexpected input shape, so now ensure_grayscale() also doesn't. Fixes issue #313.
Configuration menu - View commit details
-
Copy full SHA for 82d96f5 - Browse repository at this point
Copy the full SHA 82d96f5View commit details -
Configuration menu - View commit details
-
Copy full SHA for c389323 - Browse repository at this point
Copy the full SHA c389323View commit details -
- fixed 'str' type to 'string' - use 'string_list' type for tags list
Configuration menu - View commit details
-
Copy full SHA for 0bc92c2 - Browse repository at this point
Copy the full SHA 0bc92c2View commit details -
Configuration menu - View commit details
-
Copy full SHA for 6720372 - Browse repository at this point
Copy the full SHA 6720372View commit details
Commits on Mar 5, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 28d5eb9 - Browse repository at this point
Copy the full SHA 28d5eb9View commit details -
Configuration menu - View commit details
-
Copy full SHA for 7bd5d8e - Browse repository at this point
Copy the full SHA 7bd5d8eView commit details -
Configuration menu - View commit details
-
Copy full SHA for 6f29d39 - Browse repository at this point
Copy the full SHA 6f29d39View commit details -
Configuration menu - View commit details
-
Copy full SHA for 62f7e14 - Browse repository at this point
Copy the full SHA 62f7e14View commit details -
Configuration menu - View commit details
-
Copy full SHA for 9f0a8a1 - Browse repository at this point
Copy the full SHA 9f0a8a1View commit details -
Configuration menu - View commit details
-
Copy full SHA for 425f9f6 - Browse repository at this point
Copy the full SHA 425f9f6View commit details -
Configuration menu - View commit details
-
Copy full SHA for 311ec26 - Browse repository at this point
Copy the full SHA 311ec26View commit details
Commits on Mar 6, 2020
-
Only squeeze grayscale images.
Check if there's a single color channel before squeezing, since matplotlib wants (M, N) if single channel but (M, N, 3) for RGB. Fixes issue #320.
Configuration menu - View commit details
-
Copy full SHA for c115ee0 - Browse repository at this point
Copy the full SHA c115ee0View commit details -
Configuration menu - View commit details
-
Copy full SHA for e843259 - Browse repository at this point
Copy the full SHA e843259View commit details -
Sanitize node names used for tf loss metrics.
The node names are used for OHKM in topdown model, and so we need to ensure that the node names can be used for tf scope names. Resolves issue #319.
Configuration menu - View commit details
-
Copy full SHA for fe6945f - Browse repository at this point
Copy the full SHA fe6945fView commit details
Commits on Mar 9, 2020
-
Inference mode for learning dialog.
- only allows choice of trained configs - doesn't allow editing of config params - simplified pipeline form (without training params) - don't run unless all needed models available - validate model when user selects config with file dialog - automatically choose pipeline based on most recent trained model found in models directory for project to do: - show config options on pipeline form so that user can choose which trained model to use without going to head tab
Configuration menu - View commit details
-
Copy full SHA for c37e56f - Browse repository at this point
Copy the full SHA c37e56fView commit details -
Configuration menu - View commit details
-
Copy full SHA for 2d4c4cf - Browse repository at this point
Copy the full SHA 2d4c4cfView commit details -
Add nn.data.ensure_int function.
Converts tensor to int and scales to [0, 255] (if appropriate).
Configuration menu - View commit details
-
Copy full SHA for 21cd53f - Browse repository at this point
Copy the full SHA 21cd53fView commit details -
Configuration menu - View commit details
-
Copy full SHA for fa356ca - Browse repository at this point
Copy the full SHA fa356caView commit details -
We can't use is_floating attribute to test for floating dtype.
Configuration menu - View commit details
-
Copy full SHA for 9739b74 - Browse repository at this point
Copy the full SHA 9739b74View commit details -
Configuration menu - View commit details
-
Copy full SHA for 74327bd - Browse repository at this point
Copy the full SHA 74327bdView commit details -
Integrate tracking with topdown/bottomup inference
The tracker will be called frame by frame after predictions are converted into list of predicted instances. It will be passed the image if it's a tracker that uses the image.
Configuration menu - View commit details
-
Copy full SHA for fd5c84d - Browse repository at this point
Copy the full SHA fd5c84dView commit details -
Configuration menu - View commit details
-
Copy full SHA for 4f07a92 - Browse repository at this point
Copy the full SHA 4f07a92View commit details -
Configuration menu - View commit details
-
Copy full SHA for 540f316 - Browse repository at this point
Copy the full SHA 540f316View commit details
Commits on Mar 10, 2020
-
Wrap inference with error handling.
Directly iterating over examples in dataset means that everything stops if there's a TensorFlow error. I've added a safely_generate function which yields examples from the dataset while catching/ logging errors. This means we'll simply skip frames where we get an error, but you'll still get predictions for other frames. The function also logs progress (number of examples completed, time, and fps).
Configuration menu - View commit details
-
Copy full SHA for ea9e53d - Browse repository at this point
Copy the full SHA ea9e53dView commit details -
Configuration menu - View commit details
-
Copy full SHA for 32f1ff0 - Browse repository at this point
Copy the full SHA 32f1ff0View commit details
Commits on Mar 11, 2020
-
Configuration menu - View commit details
-
Copy full SHA for aee03b6 - Browse repository at this point
Copy the full SHA aee03b6View commit details
Commits on Mar 12, 2020
-
Configuration menu - View commit details
-
Copy full SHA for a8a7eef - Browse repository at this point
Copy the full SHA a8a7eefView commit details -
Configuration menu - View commit details
-
Copy full SHA for d6fa068 - Browse repository at this point
Copy the full SHA d6fa068View commit details -
Configuration menu - View commit details
-
Copy full SHA for 46fc4ae - Browse repository at this point
Copy the full SHA 46fc4aeView commit details -
Configuration menu - View commit details
-
Copy full SHA for 32ec659 - Browse repository at this point
Copy the full SHA 32ec659View commit details -
Bug fix for training config file list.
On first loading the dialog the file list was being updated at the wrong time and getting cleared. Now we call the update function only on subsequent shows.
Configuration menu - View commit details
-
Copy full SHA for eef6519 - Browse repository at this point
Copy the full SHA eef6519View commit details -
Configuration menu - View commit details
-
Copy full SHA for d689426 - Browse repository at this point
Copy the full SHA d689426View commit details -
Configuration menu - View commit details
-
Copy full SHA for 3c946f2 - Browse repository at this point
Copy the full SHA 3c946f2View commit details -
Configuration menu - View commit details
-
Copy full SHA for 505d34b - Browse repository at this point
Copy the full SHA 505d34bView commit details -
Configuration menu - View commit details
-
Copy full SHA for 9c628a7 - Browse repository at this point
Copy the full SHA 9c628a7View commit details -
Configuration menu - View commit details
-
Copy full SHA for b7a14cd - Browse repository at this point
Copy the full SHA b7a14cdView commit details -
Configuration menu - View commit details
-
Copy full SHA for 024b7a2 - Browse repository at this point
Copy the full SHA 024b7a2View commit details -
Configuration menu - View commit details
-
Copy full SHA for db273e1 - Browse repository at this point
Copy the full SHA db273e1View commit details -
Configuration menu - View commit details
-
Copy full SHA for e31bc1b - Browse repository at this point
Copy the full SHA e31bc1bView commit details
Commits on Mar 13, 2020
-
Configuration menu - View commit details
-
Copy full SHA for e64983a - Browse repository at this point
Copy the full SHA e64983aView commit details -
Refactor gui modules into submodules.
Created sleap.gui.dialogs and sleap.gui.widgets.
Configuration menu - View commit details
-
Copy full SHA for 465f9c9 - Browse repository at this point
Copy the full SHA 465f9c9View commit details -
Better layout/titles for datagen preview windows.
Also moved datagen preview code into separate file.
Configuration menu - View commit details
-
Copy full SHA for 4dbaa58 - Browse repository at this point
Copy the full SHA 4dbaa58View commit details -
Configuration menu - View commit details
-
Copy full SHA for e91a93d - Browse repository at this point
Copy the full SHA e91a93dView commit details -
Configuration menu - View commit details
-
Copy full SHA for 9f0a926 - Browse repository at this point
Copy the full SHA 9f0a926View commit details -
Redraw receptive field widget when changed.
When the widget moved in the form (because we added/removed a checkbox to use trained model), the old widget wasn't erased before drawing it in the new position. A call to replain() fixes this.
Configuration menu - View commit details
-
Copy full SHA for ec357b5 - Browse repository at this point
Copy the full SHA ec357b5View commit details -
Configuration menu - View commit details
-
Copy full SHA for 774408e - Browse repository at this point
Copy the full SHA 774408eView commit details -
Add training_instances property, user for training
This only includes user instances which have some visible points. Fixes issue #299.
Configuration menu - View commit details
-
Copy full SHA for 03f283e - Browse repository at this point
Copy the full SHA 03f283eView commit details -
Configuration menu - View commit details
-
Copy full SHA for 47d9073 - Browse repository at this point
Copy the full SHA 47d9073View commit details
Commits on Mar 16, 2020
-
Correctly pass full image to flow from topdown.
If the tracking module wants the full image (i.e., flow shift), then we need to save it and pass it along. This wasn't being done correctly for the topdown pipeline. Fixes issue #325.
Configuration menu - View commit details
-
Copy full SHA for f0e3d27 - Browse repository at this point
Copy the full SHA f0e3d27View commit details
Commits on Mar 17, 2020
-
Configuration menu - View commit details
-
Copy full SHA for c3bb3a5 - Browse repository at this point
Copy the full SHA c3bb3a5View commit details -
Configuration menu - View commit details
-
Copy full SHA for b320ff7 - Browse repository at this point
Copy the full SHA b320ff7View commit details -
New dialog for complex instance deleting.
The dialog allows for deleting user/predicted/all instances in specified frames/videos/tracks. The old "delete from clip" command has been modified to delete all predicted instances in the clip, regardless of whether an instance is selected. Previously it would only delete instances which matched the track of the selected instance if there was one. This was confusiong, so the current change helps address issue #312.
Configuration menu - View commit details
-
Copy full SHA for 555076d - Browse repository at this point
Copy the full SHA 555076dView commit details
Commits on Mar 18, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 3e466b9 - Browse repository at this point
Copy the full SHA 3e466b9View commit details -
Configuration menu - View commit details
-
Copy full SHA for 4fb168c - Browse repository at this point
Copy the full SHA 4fb168cView commit details -
Configuration menu - View commit details
-
Copy full SHA for f6d41ac - Browse repository at this point
Copy the full SHA f6d41acView commit details -
Configuration menu - View commit details
-
Copy full SHA for 07a3b32 - Browse repository at this point
Copy the full SHA 07a3b32View commit details -
Added mocking for multi-model inference pipelines.
- Allows for evaluation with a single model (e.g., just centroids) when ground truth data is available.
Configuration menu - View commit details
-
Copy full SHA for 372bc97 - Browse repository at this point
Copy the full SHA 372bc97View commit details -
Implement dataset filtering transformer.
- Allows for exclusion of specific elements in a dataset given a callable that evaluates a conditional on each element. - Fixes #314: bottom-up inference now properly excludes examples that when inference fails to find any local peaks.
Configuration menu - View commit details
-
Copy full SHA for 5b2fd80 - Browse repository at this point
Copy the full SHA 5b2fd80View commit details -
Configuration menu - View commit details
-
Copy full SHA for 6d232a5 - Browse repository at this point
Copy the full SHA 6d232a5View commit details -
Configuration menu - View commit details
-
Copy full SHA for 938a354 - Browse repository at this point
Copy the full SHA 938a354View commit details -
Filtering for failed centroid detection in topdown inference.
- Fixes #322: properly handles case where no centroids are detected.
Configuration menu - View commit details
-
Copy full SHA for cb02689 - Browse repository at this point
Copy the full SHA cb02689View commit details -
Normalize images for visualization
- Fixes image plotting when doing weird normalization like caffe-mode.
Configuration menu - View commit details
-
Copy full SHA for 68a2c00 - Browse repository at this point
Copy the full SHA 68a2c00View commit details -
Configuration menu - View commit details
-
Copy full SHA for e1e3ead - Browse repository at this point
Copy the full SHA e1e3eadView commit details -
Configuration menu - View commit details
-
Copy full SHA for d8070a5 - Browse repository at this point
Copy the full SHA d8070a5View commit details -
Configuration menu - View commit details
-
Copy full SHA for 91ab826 - Browse repository at this point
Copy the full SHA 91ab826View commit details -
Use Qt backend for matplotlib.
The default backend (Tk) is giving me errors related to multi- processing.
Configuration menu - View commit details
-
Copy full SHA for a9a620e - Browse repository at this point
Copy the full SHA a9a620eView commit details -
Use RLock around ffmpeg seek/read.
Without a lock we were getting errors when using viz during training. With a non-recursive lock I get a deadlock when starting inference. It appears that everything works with a recursive lock. I don't see any noticable performance cost but we should keep an eye out for this. Fixes issue #321.
Configuration menu - View commit details
-
Copy full SHA for 134e189 - Browse repository at this point
Copy the full SHA 134e189View commit details -
Configuration menu - View commit details
-
Copy full SHA for 5273c24 - Browse repository at this point
Copy the full SHA 5273c24View commit details -
Configuration menu - View commit details
-
Copy full SHA for 58082a9 - Browse repository at this point
Copy the full SHA 58082a9View commit details -
Delete outdated test and nn modules.
- Lots of these will need to be rewritten for the new format, but should be easy enough to port them from the git history later. - There's still a bit of legacy code in nn, but will refactor later.
Configuration menu - View commit details
-
Copy full SHA for 83ef7dc - Browse repository at this point
Copy the full SHA 83ef7dcView commit details -
- Commit 33339cd changed how UNet was built to ensure stride symmetry across encoder vs decoder branches, but this broke the tests. - Inspecting the output shapes closely reveals a discrepancy between the reference and the CARE implementations, so added an attribute to switch off between these. - Fixed pooling when using a stem such that stacked architectures can be correctly instantiated. - Fixed and added more tests.
Configuration menu - View commit details
-
Copy full SHA for 4cfc437 - Browse repository at this point
Copy the full SHA 4cfc437View commit details -
Configuration menu - View commit details
-
Copy full SHA for c0a6f2c - Browse repository at this point
Copy the full SHA c0a6f2cView commit details
Commits on Mar 19, 2020
-
Configuration menu - View commit details
-
Copy full SHA for 48aa9b3 - Browse repository at this point
Copy the full SHA 48aa9b3View commit details -
Create RLock only when needed.
At least for MacOS X I was getting a "too many files open" error on a dataset with very many videos even if we weren't reading frames from any of the videos.
Configuration menu - View commit details
-
Copy full SHA for f0bbd7b - Browse repository at this point
Copy the full SHA f0bbd7bView commit details -
More distinctive slider marks.
Types of vertical marks: - thick dark blue line for **suggested** frame with **user** instance, otherwise - thick light blue line for **suggested** frame with **predicted** instance, otherwise - thick black line for frame with **user** instance (whether or not it has track identity), otherwise - *thin* black line for frame with **predicted** instance **without track**, otherwise - unfilled blue rectangle for **suggested** frame without any instances. Plus horizontal bars for instances with track identities. Resolves issue #311.
Configuration menu - View commit details
-
Copy full SHA for c19f47b - Browse repository at this point
Copy the full SHA c19f47bView commit details -
Merge branch 'develop' of https://github.com/murthylab/sleap into nn-…
…pipe-integration
Configuration menu - View commit details
-
Copy full SHA for 7875d1e - Browse repository at this point
Copy the full SHA 7875d1eView commit details -
Configuration menu - View commit details
-
Copy full SHA for 383463e - Browse repository at this point
Copy the full SHA 383463eView commit details