Hi there,
I have trained a yolo11s model at 832x832 resolution, I have converted it to .hef successfully however when I go to use it on my raspi 5 with HAILO 8 I don’t get any detections. I have another version of that same dataset trained at 640x640 and that worked fine. when I converted the onnx to .hef I made sure to change the res from 640x640x3 to 832x832x3 in the yaml file. Am I missing something else?
I am using the model in the rpicam apps with the object detection pipeline for reference and have the post-process-file set to
`{
“rpicam-apps”: {
“lores”: {
“width”: 832,
“height”: 832,
“format”: “rgb”
}
},
“hailo_yolo_inference”: {
“hef_file_8L”: “/usr/share/hailo-models/yolov8s_h8l.hef”,
“hef_file_8”: “/usr/share/hailo-models/yolov11s_custom_832.hef”,
“max_detections”: 5,
“threshold”: 0.6,
“temporal_filter”: {
“tolerance”: 0.1,
“factor”: 0.75,
“visible_frames”: 6,
“hidden_frames”: 3
}
}`
so I guess the main question is, for resolutions outside of the standard 640 what special changes need to be made when converting to .hef?