detic

Detecting Twenty-thousand Classes using Image-level Supervision

Input

(Image from https://web.eecs.umich.edu/~fouhey/fun/desk/desk.jpg, credit David Fouhey)

Output

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 detic.py

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 detic.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 detic.py --video VIDEO_PATH

By adding the --model_type option, you can specify model type which is selected from "SwinB_896_4x", "R50_640_4x". (default is SwinB_896_4x)

$ python3 detic.py --model_type SwinB_896_4x

By adding the --vocabulary option, you can specify the model's vocabulary which is selected from "lvis", "in21k". (default is lvis)

$ python3 detic.py --vocabulary lvis

By adding the --detection_width option, you can specify the model's input width to increase inference speed

$ python3 detic.py --detection_width 320

Reference

• Detecting Twenty-thousand Classes using Image-level Supervision

Framework

Pytorch

Model Format

ONNX opset=11

Netron

• Detic_C2_SwinB_896_4x_IN-21K+COCO_lvis.onnx.prototxt
• Detic_C2_SwinB_896_4x_IN-21K+COCO_in21k.onnx.prototxt
• Detic_C2_R50_640_4x_lvis.onnx.prototxt
• Detic_C2_R50_640_4x_in21k.onnx.prototxt