I am using YOLOv7 trained on custom dataset, and using the model weights after converting to ONNX for CVAT annotation. There it's only predicting few small objects. Is there any limit on small object detection in YOLOv7. In every image, it's only predicting first 3-4 characters in the plate.
As explained in the question Is there any limit to detected objects in YOLOv7?, while converting to ONNX, I modified the 'topk-all' to 400, but still no change in result
Resolved it by increasing the epoch.