facebookresearch / facebookresearch/SlowFast

No action detections in output video after running run_net.py for demo

Open
#696 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
7.4k
Forks
1.3k
PR merge metrics
No merged PRs in 30d

Description

I trained a model with AVA format custom dataset after 20 epochs, during training it seemed to work well and even there were some bugs and I thought I had fixed them. I used one of checkpoint models ``checkpoint_16.pyth`` and ran the command ``python tools/run_net.py --cfg /home/young/Projects/Slowfast/SlowFast/demo/AVA/slowfast_demo.yaml``,the output video did not have any action detections, it seemed there were no bugs, do i need to train Detectron2 for object detection in my custom dataset ? some information like these:

[02/7 10:23][INFO] predictor.py: 185: Initialized Detectron2 Object Detection Model.
[02/7 10:23][INFO] detection_checkpoint.py: 38: [DetectionCheckpointer] Loading from detectron2://COCO-Detection/faster_rcnn_R_50_FPN_3x/137849458/model_final_280758.pkl ...
[02/7 10:23][INFO] file_io.py: 798: URL https://dl.fbaipublicfiles.com/detectron2/COCO-Detection/faster_rcnn_R_50_FPN_3x/137849458/model_final_280758.pkl cached in /home/yg/.torch/iopath_cache/detectron2/COCO-Detection/faster_rcnn_R_50_FPN_3x/137849458/model_final_280758.pkl
[02/7 10:23][INFO] checkpoint.py: 150: [Checkpointer] Loading from /home/yg/.torch/iopath_cache/detectron2/COCO-Detection/faster_rcnn_R_50_FPN_3x/137849458/model_final_280758.pkl ...
[02/7 10:23][INFO] detection_checkpoint.py: 76: Reading a file from 'Detectron2 Model Zoo'
[02/7 10:23][INFO] predictor.py: 45: Start loading model weights.
[02/7 10:23][INFO] checkpoint.py: 222: Loading network weights from ./checkpoints/checkpoint_16_v1.pyth.
missing keys: []
unexpected keys: []
[02/7 10:24][INFO] predictor.py: 47: Finish loading model weights
/home/yg/anaconda3/envs/Slowfast/lib/python3.9/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at ../aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
7it [02:24, 20.59s/it]
[02/7 10:41][INFO] demo_net.py: 119: Finish demo in: 144.79783868789673

Thanks you!!

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.