Demo Usage

Currently the demo supports visualization for:

• Image Folder: A set of frames that were decoded from a given video.
• Video: I only tested .mp4, but other video format should be OK.

Inference on a image folder

The command line should be like this:

    python demo/demo.py ${METHOD} ${CONFIG_FILE} ${CHECKPOINT_FILE} [--visualize-path ${IMAGE-FOLDER}] [--suffix ${IMAGE_SUFFIX}][--output-folder ${FOLDER}] [--output-video]

Example:

    python demo/demo.py base configs/vid_R_101_C4_1x.yaml R_101.pth --suffix ".JPEG"\
        --visualize-path datasets/ILSVRC2015/Data/VID/val/ILSVRC2015_val_00003001 \
        --output-folder visualization [--output-video]

This will generate visualization result using single frame baseline with ResNet-101 backbone. And the results, images with generated bboxes, are saved in folder visualization.

Please note that:

1. If your want to use other methods like MEGA, FGFA, please change METHOD base to mega or fgfa. Currently all methods support visualization, see demo.py for more information about using other methods.
2. Don't forget to modify CONFIG_FILE and CHECKPOINT_FILE accordingly!
3. Add --output-video to generate video instead of set of images, the video is encoded at 25 fps by default.
4. If you want to visualize your own image folder, please make sure that the name of your images is like XXXXXX.jpg. XXXXXX is the frame number of current frame, e.g., 000000 is the first frame. .jpg could be replaced by other common image suffix like .png, which could be specified by --suffix.

Inference on a video

The command line should be like this:

    python demo/demo.py ${METHOD} ${CONFIG_FILE} ${CHECKPOINT_FILE} --video [--visualize-path ${VIDEO-NAME}] [--output-folder ${FOLDER}] [--output-video]

Example:

    python demo/demo.py base configs/vid_R_101_C4_1x.yaml R_101.pth --video \
        --visualize-path datasets/ILSVRC2015/Data/VID/snippets/val/ILSVRC2015_val_00003001.mp4 \
        --output-folder visualization [--output-video]

This will generate visualization result using single frame baseline with ResNet-101 backbone. And the results, images with generated bboxes, are saved in folder visualization.

Please note that:

1. All you should know about has given above.

Misc

Nothing more is needed?