what is this anchor about forge HOT 2 OPEN

It's a set of predefined bounding boxes at different (fixed) positions and sizes. Rather than directly predicting bounding box coordinates, this version of YOLO predicts how each of these predefined bounding boxes (anchors) are moved and scaled to fit objects in the image.

from forge.

Hi @hollance,
I'm currently working on an ONNX converter for the MLPack framework (a C++ machine learning framework). I've successfully developed an initial prototype for the ONNX-MLPack converter, which works fine with the yolov2.onnx model you provided since it has a simple graph. Now, I'm trying to test my converter with the yolov3.onnx models I found online, but all of them seem to have very complex graphs.

these are some of the yolov3.onnx models i found online but all have very complex graph
https://github.com/onnx/models/tree/main/validated/vision/object_detection_segmentation/yolov3
https://github.com/onnx/models/tree/main/validated/vision/object_detection_segmentation/tiny-yolov3

if you have worked on yolov3.onnx then could you please provide some yolov3.onnx model with simpler graph

from forge.