Quick fix for data pipeline

README.md

@@ -29,19 +29,21 @@ Paper: [High resolution](https://drive.google.com/file/d/0B2dg5RanEUBQRkJYZDc1Mm
 You may choose the ideal version to use
 
 - Clone [this](https://github.com/eborboihuc/Deep360Pilot-CVPR17) repo and [another](https://github.com/yenchenlin/Deep360Pilot-optical-flow) for formating the input data:
+
 ```bash
-git clone git@github.com:eborboihuc/Deep360Pilot-CVPR17.git
+git clone http://github.com/eborboihuc/Deep360Pilot-CVPR17.git
+
 cd Deep360Pilot/misc
 
-git clone git@github.com:yenchenlin/Deep360Pilot-optical-flow.git
+git clone http://github.com/yenchenlin/Deep360Pilot-optical-flow.git
 ```
 - Download our [dataset](#dataset) and [pre-trained model](#pre-trained-model)
 
 After run the scripts
 ```bash
 python require.py
 ```
-You can download our model and dataset and place it under `./checkpoint` and `./data`, respectively.
+Please download our model and dataset and place it under `./checkpoint` and `./data`, respectively.
 
 
 # Usage
@@ -73,13 +75,21 @@ Please download the trained model [here](https://drive.google.com/uc?export=down
 You can use `--model {model_path}` in `main.py` to load the model. 
 
 ## Dataset
-If you want to test on our dataset, please download the dataset [here](https://drive.google.com/uc?export=download&id=0B9wE6h4m--wjWnF3LV9WUXdZMzA) and place it under `./data`.
-You can use code from [here](https://github.com/yenchenlin/Deep360Pilot-optical-flow) to convert it to our input format.
 
-## Author
+### Pipeline testing
+We provide a small testing clip-based datafile. Please download it [here](https://drive.google.com/uc?export=download&id=0B9wE6h4m--wjaTNPYUk4NkM0UDA). And you can use this toy datafile to go though our data process pipeline.
+
+### Testing on our *batch-based dataset* for accuracy and smoothness
+If you want to reproduce the results on our dataset, please download the dataset [here](https://drive.google.com/uc?export=download&id=0B9wE6h4m--wjWnF3LV9WUXdZMzA) and place it under `./data`.
+
+### Testing on our *clip-based dataset* for generating trajectories
+Please download the *clip-based dataset* [here](https://drive.google.com/uc?export=download&id=0B9wE6h4m--wjZzJkZnNLZW1BNE0)
+And then use code from [here](https://github.com/yenchenlin/Deep360Pilot-optical-flow) to convert it to our input format.
+
+# Author
 Hou-Ning Hu / [@eborboihuc](https://eborboihuc.github.io/) and Yen-Chen Lin / [@yenchenlin](https://yclin.me)
 
-## Cite
+# Cite
 If you find our code useful for your research, please cite
 ```bibtex
 @InProceedings{Hu_2017_CVPR,

misc/README.md

@@ -1,16 +1,40 @@
 # 360 videos data pipeline
 
 ## How to Run?
-1. Make sure `dataset.csv` is placed in the repository, which contains the metadata of our dataset. And also place `OF_360` repository in the current location.
 
-2. Parse `dataset.csv` to generate `metadata.npy`, run:
+*NOTE*: You can use this [example](https://drive.google.com/uc?export=download&id=0B9wE6h4m--wjaTNPYUk4NkM0UDA) file to go through data pipeline
+
+1. Make sure `dataset.csv` is placed in the repository, which contains the metadata of our dataset. 
+   And also place `Deep360Pilot-optical-flow` repository in the current location.
+
+2. Make sure the generated feature `roisavg.npy`, `roislist.npy` and every video frames placed at `data/feature_{DOMAIN}_{N_BOXES}boxes` and `data/frame_{DOMAIN}`
+
+3. Execute `run.sh {DOMAIN} {N_BOXES}` once and you can get the ideal data format under `data/feature_{DOMAIN}_{N_BOXES}boxes`
+
+  `./run.sh test 16`
+
+
+
+## Details inside run.sh
+
+1. Parse `dataset.csv` to generate `metadata.npy`, run:
 
   `python parse_csv.py`
 
-3. Split raw label according to each clip which has length 50, run:
+2. Split raw label according to each clip which has length 50, run:
 
   `python parse_raw_label.py`
 
+3. Deep360Pilot-Optical-Flow will generate ideal data input format and also hof.npy files, see `run.sh`:
+
+  ```bash
+  python Deep360Pilot-optical-flow/compute_video_flow.py -d $DOMAIN
+  python Deep360Pilot-optical-flow/divide_area_motion_salient_boxes.py -d $DOMAIN -b $N_BOXES
+  python Deep360Pilot-optical-flow/pruned_box_features.py -d $DOMAIN -b $N_BOXES
+  python Deep360Pilot-optical-flow/get_motion_features.py -d $DOMAIN -b $N_BOXES
+  python Deep360Pilot-optical-flow/hist_of_opt_flow/get_hof.py -d $DOMAIN -b $N_BOXES
+  ```
+
 4. Generate batches, run:
 
   `python gen_batch.py`

misc/config.py

@@ -1,5 +1,5 @@
 import os
 
 ROOT_PATH = os.path.abspath(os.path.join(os.getcwd(), os.path.pardir))
-FEATURE_PATH = os.path.join(ROOT_PATH, 'newData')
-BATCH_PATH = os.path.join(ROOT_PATH, 'dataset')
+BATCH_PATH = os.path.abspath(os.path.join(os.getcwd(), 'dataset'))
+FEATURE_PATH = os.path.join(ROOT_PATH, 'data')

misc/run.sh

@@ -8,22 +8,44 @@ fi
 DOMAIN=${1}
 N_BOXES=${2}
 
+# Function
+echo_time () {
+    echo [$(date)] ${1}
+}
+
+# Mkdir
+_dirs=('dataset' 'metadata')
+for dirname in "${_dirs[@]}"
+do
+    if [ ! -d ${dirname} ]; then
+        echo_time "mkdir ${dirname}"
+        mkdir ${dirname}
+    else
+        echo_time "${dirname} exist"
+    fi
+done
+
 python parse_csv.py -d $DOMAIN
 
 # """ In case you want to label new domain """
-#sudo python parse_raw_label.py -d $DOMAIN -b $N_BOXES
-#sudo python one_hot_convert.py -m data -d $DOMAIN -b $N_BOXES
-
+#python parse_raw_label.py -d $DOMAIN -b $N_BOXES
+#python one_hot_convert.py -m data -d $DOMAIN -b $N_BOXES
 
-python OF_360/divide_area_motion_salient_boxes.py -d $DOMAIN -b $N_BOXES
-python OF_360/pruned_box_features.py -d $DOMAIN -b $N_BOXES
-python OF_360/get_motion_features.py -d $DOMAIN -b $N_BOXES
-python OF_360/hist_of_opt_flow/get_hof.py -d $DOMAIN -b $N_BOXES
+# """ Convert flow and apply to features """
+echo_time "python Deep360Pilot-optical-flow/compute_video_flow.py -d $DOMAIN"
+python Deep360Pilot-optical-flow/compute_video_flow.py -d $DOMAIN
+echo_time "python Deep360Pilot-optical-flow/divide_area_motion_salient_boxes.py -d $DOMAIN -b $N_BOXES"
+python Deep360Pilot-optical-flow/divide_area_motion_salient_boxes.py -d $DOMAIN -b $N_BOXES
+echo_time "python Deep360Pilot-optical-flow/pruned_box_features.py -d $DOMAIN -b $N_BOXES"
+python Deep360Pilot-optical-flow/pruned_box_features.py -d $DOMAIN -b $N_BOXES
+echo_time "python Deep360Pilot-optical-flow/get_motion_features.py -d $DOMAIN -b $N_BOXES"
+python Deep360Pilot-optical-flow/get_motion_features.py -d $DOMAIN -b $N_BOXES
+echo_time "python Deep360Pilot-optical-flow/hist_of_opt_flow/get_hof.py -d $DOMAIN -b $N_BOXES"
+python Deep360Pilot-optical-flow/hist_of_opt_flow/get_hof.py -d $DOMAIN -b $N_BOXES
 
 
 # """ In case you need to generate test batches """
-#cd batch_data/$DOMAIN
-#cd ../..
+#cd data/${DOMAIN}_${N_BOXES}
 #python mkdir.py -d $DOMAIN
 #python gen_batch.py -d $DOMAIN -b $N_BOXES
 #python test_gen_batch.py -d $DOMAIN -b $N_BOXES

require.py

@@ -14,4 +14,18 @@
 if os.path.isdir(os.path.join(root, 'misc/data')) == False:
     os.symlink(os.path.join(root, 'data'), os.path.join(root, 'misc/data'))
 
+cmd='wget --no-check-certificate "https://drive.google.com/uc?export=download&id={ID}" -O {LOCAL_FILE_NAME}'
+show='https://drive.google.com/uc?export=download&id={ID} and place it at {LOCAL_FILE_NAME}'
 
+print "Please Download these files:"
+
+print show.format(**{'ID': '0B9wE6h4m--wjNWdFbnVYbG9kNm8', 'LOCAL_FILE_NAME': 'checkpoint/Deep360Pilot-model.zip'})
+#os.system(cmd.format(**{'ID': '0B9wE6h4m--wjNWdFbnVYbG9kNm8', 'LOCAL_FILE_NAME': 'checkpoint/Deep360Pilot-model.zip'}))
+print show.format(**{'ID': '0B9wE6h4m--wjaTNPYUk4NkM0UDA', 'LOCAL_FILE_NAME': 'data/Deep360Pilot-test.zip'})
+#os.system(cmd.format(**{'ID': '0B9wE6h4m--wjaTNPYUk4NkM0UDA', 'LOCAL_FILE_NAME': 'data/Deep360Pilot-test.zip'}))
+print show.format(**{'ID': '0B9wE6h4m--wjWnF3LV9WUXdZMzA', 'LOCAL_FILE_NAME': 'data/Deep360Pilot-feature.zip'})
+#os.system(cmd.format(**{'ID': '0B9wE6h4m--wjWnF3LV9WUXdZMzA', 'LOCAL_FILE_NAME': 'data/Deep360Pilot-feature.zip'}))
+print show.format(**{'ID': '0B9wE6h4m--wjZzJkZnNLZW1BNE0', 'LOCAL_FILE_NAME': 'data/Deep360Pilot-batch-feature.zip'})
+#os.system(cmd.format(**{'ID': '0B9wE6h4m--wjZzJkZnNLZW1BNE0', 'LOCAL_FILE_NAME': 'data/Deep360Pilot-batch-feature.zip'}))
+print show.format(**{'ID': '0B9wE6h4m--wjSXVYblBWNktacFk', 'LOCAL_FILE_NAME': 'data/Deep360Pilot-label.zip'})
+#os.system(cmd.format(**{'ID': '0B9wE6h4m--wjSXVYblBWNktacFk', 'LOCAL_FILE_NAME': 'data/Deep360Pilot-label.zip'}))