Implement Yolo (CornerNet and/or Saccade)

[AlexeyAB]

• MS COCO state-of-art accuracy: https://paperswithcode.com/sota/object-detection-on-coco

• CornerNet: Detecting Objects as Paired Keypoints: https://arxiv.org/abs/1808.01244v2

• CornerNet-Lite: Efficient Keypoint Based Object Detection: https://arxiv.org/abs/1904.08900v1 https://www.groundai.com/project/cornernet-lite-efficient-keypoint-based-object-detection/1

  (1) We propose CornerNet-Saccade and CornerNet-Squeeze, two novel approaches to improving the efficiency of keypoint-based object detection; (2) On COCO, we improve the efficiency of state-of-the-art keypoint based detection by 6 fold and the AP from 42.2% to 43.2%, (3) On COCO, we improve both the accuracy and efficiency of state-of-the art real-time object detection (to 34.4% at 30ms from 33.0% at 39ms of YOLOv3).

• CornerNet-Saccade and CornerNet-Squeeze with PyTorch: https://github.com/princeton-vl/CornerNet-Lite

---

---

[AlexeyAB]

https://arxiv.org/pdf/1904.07850v2.pdf

[AlexeyAB]

Added model: https://github.com/AlexeyAB/darknet/files/4007007/tiny_v3_pan3_CenterNet_Gaus_ae_mosaic_scale_iouthresh_mosaic.txt

Which generates bboxes from three points of the object (top-left Corner, right-bottom Corner, Center), and then merges them using the NMS.

That should be trained by using command ./darknet detector train data/self_driving.data tiny_v3_pan3_CenterNet_Gaus_ae_mosaic_scale_iouthresh_mosaic.cfg yolov3-tiny.conv.14 -map

Whole comparison: https://github.com/AlexeyAB/darknet/issues/3114#issuecomment-494148968

---

Model (cfg & weights) network size = 544x544	Training chart	Validation video	BFlops	Inference time RTX2070, ms	mAP, %
yolo_v3_tiny_pan3.cfg.txt and weights-file Features: PAN3, AntiAliasing, Assisted Excitation, scale_x_y, Mixup, GIoU		video	14	8.5 ms	67.3%
yolo_v3_tiny_pan5 matrix_gaussian_GIoU aa_ae_mixup_new.cfg.txt and weights-file Features: MatrixNet, Gaussian-yolo + GIoU, PAN5, IoU_thresh, Deformation-block, Assisted Excitation, scale_x_y, Mixup, 512x512, use -thresh 0.6		video	30	31 ms	64.6%
yolo_v3_spp_pan_scale.cfg.txt and weights-file		video	137	33.8 ms	60.4%
tiny_v3_pan3_CenterNet_Gaus ae_mosaic_scale_iouthresh mosaic.txt and weights-file		video	25	14.5 ms	57.9%
yolov3-tiny_3l.cfg.txt (common old model) and weights-file		video	12	5.6 ms	46.8%
yolo_v3_tiny.cfg.txt (common old model) and weights-file		video	9	5.0 ms	32.3%
				-	-
				-	-