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.

Sign up for GitHub

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

[Docs] Highlight loading checkpoints directly from url #339

Merged
merged 2 commits into from
Nov 11, 2020
Merged

[Docs] Highlight loading checkpoints directly from url #339

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bd23ad6
highlight function to load checkpoint from url
dreamerlin Nov 11, 2020
2094e3a
highlight function to load checkpoint from url
dreamerlin Nov 11, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 29 additions & 9 deletions demo/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,8 @@ Optional arguments:

Examples:

Assume that you are located at `$MMACTION2` and have already downloaded the checkpoints to the directory `checkpoints/`
Assume that you are located at `$MMACTION2` and have already downloaded the checkpoints to the directory `checkpoints/`,
or use checkpoint url from `configs/` to directly load corresponding checkpoint, which will be automatically saved in `$HOME/.cahe/torch/checkpoints`.

1. Recognize a video file as input by using a TSN model on cuda by default.

Expand All @@ -38,15 +39,24 @@ Assume that you are located at `$MMACTION2` and have already downloaded the chec
demo/demo.mp4 demo/label_map.txt
```

2. Recognize a list of rawframes as input by using a TSN model on cpu.
2. Recognize a video file as input by using a TSN model on cuda by default, loading checkpoint from url.

```shell
# The demo.mp4 and label_map.txt are both from Kinetics-400
python demo/demo.py configs/recognition/tsn/tsn_r50_video_inference_1x1x3_100e_kinetics400_rgb.py \
https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_1x1x3_100e_kinetics400_rgb/tsn_r50_1x1x3_100e_kinetics400_rgb_20200614-e508be42.pth \
demo/demo.mp4 demo/label_map.txt
```

3. Recognize a list of rawframes as input by using a TSN model on cpu.

```shell
python demo/demo.py configs/recognition/tsn/tsn_r50_inference_1x1x3_100e_kinetics400_rgb.py \
checkpoints/tsn_r50_1x1x3_100e_kinetics400_rgb_20200614-e508be42.pth \
PATH_TO_FRAMES/ LABEL_FILE --use-frames --device cpu
```

3. Recognize a video file as input by using a TSN model and then generate an mp4 file.
4. Recognize a video file as input by using a TSN model and then generate an mp4 file.

```shell
# The demo.mp4 and label_map.txt are both from Kinetics-400
Expand All @@ -55,15 +65,15 @@ Assume that you are located at `$MMACTION2` and have already downloaded the chec
demo/demo.mp4 demo/label_map.txt --out-filename demo/demo_out.mp4
```

4. Recognize a list of rawframes as input by using a TSN model and then generate a gif file.
5. Recognize a list of rawframes as input by using a TSN model and then generate a gif file.

```shell
python demo/demo.py configs/recognition/tsn/tsn_r50_inference_1x1x3_100e_kinetics400_rgb.py \
checkpoints/tsn_r50_1x1x3_100e_kinetics400_rgb_20200614-e508be42.pth \
PATH_TO_FRAMES/ LABEL_FILE --use-frames --out-filename demo/demo_out.gif
```

5. Recognize a video file as input by using a TSN model, then generate an mp4 file with a given resolution and resize algorithm.
6. Recognize a video file as input by using a TSN model, then generate an mp4 file with a given resolution and resize algorithm.

```shell
# The demo.mp4 and label_map.txt are both from Kinetics-400
Expand All @@ -83,7 +93,7 @@ Assume that you are located at `$MMACTION2` and have already downloaded the chec
--out-filename demo/demo_out.mp4
```

6. Recognize a video file as input by using a TSN model, then generate an mp4 file with a label in a red color and 10px fontsize.
7. Recognize a video file as input by using a TSN model, then generate an mp4 file with a label in a red color and 10px fontsize.

```shell
# The demo.mp4 and label_map.txt are both from Kinetics-400
Expand All @@ -93,7 +103,7 @@ Assume that you are located at `$MMACTION2` and have already downloaded the chec
--out-filename demo/demo_out.mp4
```

7. Recognize a list of rawframes as input by using a TSN model and then generate an mp4 file with 24 fps.
8. Recognize a list of rawframes as input by using a TSN model and then generate an mp4 file with 24 fps.

```shell
python demo/demo.py configs/recognition/tsn/tsn_r50_inference_1x1x3_100e_kinetics400_rgb.py \
Expand All @@ -119,7 +129,8 @@ Optional arguments:

Examples:

Assume that you are located at `$MMACTION2` and have already downloaded the checkpoints to the directory `checkpoints/`
Assume that you are located at `$MMACTION2` and have already downloaded the checkpoints to the directory `checkpoints/`,
or use checkpoint url from `configs/` to directly load corresponding checkpoint, which will be automatically saved in `$HOME/.cahe/torch/checkpoints`.

1. Recognize the action from web camera as input by using a TSN model on cpu, averaging the score per 5 times
and outputting result labels with score higher than 0.2.
Expand All @@ -130,7 +141,16 @@ Assume that you are located at `$MMACTION2` and have already downloaded the chec
--threshold 0.2 --device cpu
```

2. Recognize the action from web camera as input by using a I3D model on gpu by default, averaging the score per 5 times
2. Recognize the action from web camera as input by using a TSN model on cpu, averaging the score per 5 times
and outputting result labels with score higher than 0.2, loading checkpoint from url.

```shell
python demo/webcam_demo.py configs/recognition/tsn/tsn_r50_video_inference_1x1x3_100e_kinetics400_rgb.py \
https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_1x1x3_100e_kinetics400_rgb/tsn_r50_1x1x3_100e_kinetics400_rgb_20200614-e508be42.pth \
demo/label_map.txt --average-size 5 --threshold 0.2 --device cpu
```

3. Recognize the action from web camera as input by using a I3D model on gpu by default, averaging the score per 5 times
and outputting result labels with score higher than 0.2.

```shell
Expand Down
4 changes: 2 additions & 2 deletions demo/demo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@
def parse_args():
parser = argparse.ArgumentParser(description='MMAction2 demo')
parser.add_argument('config', help='test config file path')
parser.add_argument('checkpoint', help='checkpoint file')
parser.add_argument('checkpoint', help='checkpoint file/url')
parser.add_argument('video', help='video file/url or rawframes directory')
parser.add_argument('label', help='label file')
parser.add_argument(
Expand Down Expand Up @@ -128,7 +128,7 @@ def main():
args = parse_args()
# assign the desired device.
device = torch.device(args.device)
# build the recognizer from a config file and checkpoint file
# build the recognizer from a config file and checkpoint file/url
model = init_recognizer(
args.config,
args.checkpoint,
Expand Down
2 changes: 1 addition & 1 deletion mmaction/apis/inference.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@ def init_recognizer(config,
Args:
config (str | :obj:`mmcv.Config`): Config file path or the config
object.
checkpoint (str | None, optional): Checkpoint path. If left as None,
checkpoint (str | None, optional): Checkpoint path/url. If set to None,
the model will not load any weights. Default: None.
device (str | :obj:`torch.device`): The desired device of returned
tensor. Default: 'cuda:0'.
Expand Down