mmfashion

MMFashion

Input

(Image from https://github.com/open-mmlab/mmfashion/blob/master/demo/imgs/01_4_full.jpg)

Shape : (1, 3, height, width)

Output

• bboxes shape : (objects, bbox)
• labels shape : (objects)
• masks shape : (objects, 28, 28)
• bbox : (left, top, right, bottom, probability)
• probability : [0.0,1.0]

Category

CATEGORY = (
    'top', 'skirt', 'leggings', 'dress', 'outer', 'pants', 'bag',
    'neckwear', 'headwear', 'eyeglass', 'belt', 'footwear', 'hair',
    'skin', 'face'
)

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 mmfashion.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 mmfashion.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 mmfashion.py --video VIDEO_PATH

By specifying the 'large' or 'small' (architecture of the u2net model) with the -pp option, the background of the input image would be removed before inference.
This process improves the accuracy of detection.

$ python3 mmfashion.py -pp large

Reference

• MMDetection
• MMFashion
• OTEDetection

Framework

ONNX Runtime

Model Format

ONNX opset=10

Netron

mask_rcnn_r50_fpn_1x.onnx.prototxt