Object_Detection: NanoDet (GSOC)

[Sidd1609]

Updated files for object detection using YOLOX, files uploaded for review:

• Model files (.onnx format FP16)
• Model: Nanodet-plus-m-1.5x-416
• README
• LICENSE (Apache 2.0)
• demo.py (with image and webcam inference options)
• NanoDet.py (class NanoDet)
• examples (test_data: test images, results: inferred results)
• COCO class names (coco.names)

COMPLETED:

• [x] Model upload
• [x] demo file
• [x] NanoDet file
• [x] Example inference (image and webcam)

TO DO:

• [x] Model quantization

[fengyuentau]

This PR is for NanoDet, so please remove the object_detection_yolox directory.

[Sidd1609]

removed yolox directory

[Sidd1609]

Yes, resolving the changes

On Sun, Sep 4, 2022 at 7:38 AM Yuantao Feng @.***> wrote:

@.**** commented on this pull request.

In models/object_detection_nanodet/NanodetPlus.py https://github.com/opencv/opencv_zoo/pull/87#discussion_r962228626:

•    with open('coco.names', 'rt') as f:
  

•        self.classes = f.read().rstrip('\n').split('\n')
  

Lets keep all classes here in this file like this https://github.com/open-mmlab/mmdetection/blob/df28da98926bc410e16bed1e9fc7d425d9a89495/mmdet/datasets/coco.py#L25-L38

In models/object_detection_nanodet/NanodetPlus.py https://github.com/opencv/opencv_zoo/pull/87#discussion_r962228759:

•    top, left, newh, neww = 0, 0, self.image_shape[0], self.image_shape[1]
  

•    if keep_ratio and srcimg.shape[0] != srcimg.shape[1]:
  

•        hw_scale = srcimg.shape[0] / srcimg.shape[1]
  

•        if hw_scale > 1:
  

•            newh, neww = self.image_shape[0], int(self.image_shape[1] / hw_scale)
  

•            img = cv2.resize(srcimg, (neww, newh), interpolation=cv2.INTER_AREA)
  

•            left = int((self.image_shape[1] - neww) * 0.5)
  

•            img = cv2.copyMakeBorder(img, 0, 0, left, self.image_shape[1] - neww - left, cv2.BORDER_CONSTANT,
  

•                                     value=0)  # add border
  

•        else:
  

•            newh, neww = int(self.image_shape[0] * hw_scale), self.image_shape[1]
  

•            img = cv2.resize(srcimg, (neww, newh), interpolation=cv2.INTER_AREA)
  

•            top = int((self.image_shape[0] - newh) * 0.5)
  

•            img = cv2.copyMakeBorder(img, top, self.image_shape[0] - newh - top, 0, 0, cv2.BORDER_CONSTANT, value=0)
  

•    else:
  

•        img = cv2.resize(srcimg, self.image_shape, interpolation=cv2.INTER_AREA)
  

•    blob = self.pre_process(img)
  

Can we reuse the code from yolox for letterbox preprocessing?

In models/object_detection_nanodet/NanodetPlus.py https://github.com/opencv/opencv_zoo/pull/87#discussion_r962228839:

•                                     value=0)  # add border
  

•        else:
  

•            newh, neww = int(self.image_shape[0] * hw_scale), self.image_shape[1]
  

•            img = cv2.resize(srcimg, (neww, newh), interpolation=cv2.INTER_AREA)
  

•            top = int((self.image_shape[0] - newh) * 0.5)
  

•            img = cv2.copyMakeBorder(img, top, self.image_shape[0] - newh - top, 0, 0, cv2.BORDER_CONSTANT, value=0)
  

•    else:
  

•        img = cv2.resize(srcimg, self.image_shape, interpolation=cv2.INTER_AREA)
  

•    blob = self.pre_process(img)
  

•    self.net.setInput(blob)
  

•    outs = self.net.forward(self.net.getUnconnectedOutLayersNames())
  

•    det_bboxes, det_conf, det_classid = self.post_process(outs)
  

•    ratioh,ratiow = srcimg.shape[0]/newh,srcimg.shape[1]/neww
  

•    return left, top, ratioh, ratiow, det_bboxes, det_conf, det_classid
  

I propose to have the same formatting of results of yolox.

In models/object_detection_nanodet/coco.names https://github.com/opencv/opencv_zoo/pull/87#discussion_r962228905:

@@ -0,0 +1,80 @@ +person

This file is not needed if we keep these classes in the wrapper class.

In models/object_detection_nanodet/demo.py https://github.com/opencv/opencv_zoo/pull/87#discussion_r962228939:

•    if(Video_save):
  

•        clip = moviepy.video.io.ImageSequenceClip.ImageSequenceClip(frame_list, fps=fps)
  

•        clip.write_videofile('Webcam_result.mp4')
  

Use VideoWriter instead.

— Reply to this email directly, view it on GitHub https://github.com/opencv/opencv_zoo/pull/87#pullrequestreview-1095584259, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALTC7V6AFWSL5FZ72IKDEKDV4QAANANCNFSM57AGTIAA . You are receiving this because you authored the thread.Message ID: @.***>

[fengyuentau]

@Sidd1609 could you squash commits into one? I am going to work on this pr and merge it.

[Sidd1609]

Changes have been made for nanodet And yolox commits have been updated in separate branch

On Fri, Oct 21, 2022 at 12:37 PM Yuantao Feng @.***> wrote:

@Sidd1609 https://github.com/Sidd1609 could you squash commits into one? I am going to work on this pr and merge it.

— Reply to this email directly, view it on GitHub https://github.com/opencv/opencv_zoo/pull/87#issuecomment-1286548654, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALTC7V3SWF7K6DDP2F5W6OTWEI6MHANCNFSM57AGTIAA . You are receiving this because you were mentioned.Message ID: @.***>

[fengyuentau]

close in favor of https://github.com/opencv/opencv_zoo/pull/103