Skip to content
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

can pocketflow be used for object detection model #44

Closed
AllenDun opened this issue Nov 12, 2018 · 16 comments
Closed

can pocketflow be used for object detection model #44

AllenDun opened this issue Nov 12, 2018 · 16 comments
Assignees
Labels
enhancement New feature or request

Comments

@AllenDun
Copy link

thanks for the great work. Many tasks in industry scene are object detections. Can pocketflow be used for object detection model, such as ssd, faster rcnn?

@as754770178
Copy link

as754770178 commented Nov 12, 2018

PocketFlow does not support object detection models (YOLO / Faster R-CNN) for the moment. We are working on this right now. #35

@jiaxiang-wu jiaxiang-wu self-assigned this Nov 12, 2018
@jiaxiang-wu jiaxiang-wu added the enhancement New feature or request label Nov 12, 2018
@jiaxiang-wu
Copy link
Contributor

Enhancement required: add support for object detection models.

@jiaxiangxu
Copy link

Excuse me, what is the fundamental difference between object detection models and image classification models such that the current PocketFlow cannot support object detection models?

@jiaxiang-wu
Copy link
Contributor

jiaxiang-wu commented Nov 20, 2018

@jiaxiangxu
Some object detection models, e.g. Faster R-CNN and R-FCN, require multiple training stages (RPN -> detection model -> RPN with shared conv. layers -> detection model with shared conv. layers). This is not yet supported in PocketFlow.

@jiaxiangxu
Copy link

jiaxiangxu commented Nov 20, 2018

thank you, then why present PocketFlow cannot support SSD and YOLO?

@jiaxiang-wu
Copy link
Contributor

@jiaxiangxu
We have not tested on SSD on YOLO yet, so we are not sure whether PocketFlow can be directly used for these two models.

@jiaxiang-wu
Copy link
Contributor

SSD models have been supported in the latest master branch.

@as754770178
Copy link

sorry, I don't find the document about pruning on SSD. Can you tell me?

@jiaxiang-wu
Copy link
Contributor

@as754770178 Documentation on SSD models is not yet ready. We will publish the corresponding documentation once the channel pruning component can work as expected on SSD models.

@as754770178
Copy link

The compression on SSD is only quantization? Is there document about quantization on SSD?

@jiaxiang-wu
Copy link
Contributor

Current code only supports uniform quantization, but we will release the support for channel pruning shortly. The document for uniform quantization and channel pruning will be released together.

@as754770178
Copy link

When you release the document, I hope you can inform me.

@jiaxiang-wu
Copy link
Contributor

@as754770178 No problem.

@howtocodewang
Copy link

@jiaxiang-wu Can you tell me when the document for object detection network compression will be released.

@sourman
Copy link

sourman commented Apr 12, 2019

@jiaxiang-wu I am curious if channel pruning has been released. I would like to try it for Mobilenet SSD architecture

@smohan10
Copy link

I'm curious to know if channel pruning can be done on SSD model on COCO dataset using Pocketflow. Has anyone tried it and were successful?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

7 participants