Then I've saved the result as .pb file, which seems to work on darkflow environment, but it doesn't predict anything when it's integrated into OpenCVForUnity.
This integration it's done by setting the dnn net as _Net yoloNet = readNet("pb_pathfile/yolo-voc-3c.pb") , passing the blob input to the net (yoloNet.setinput(blob)) and doing its forward, it returns no boundingbox outputs. But there must being something wrong on that transference since I know the yolo-voc-3c.pb properly works, because it returns both bounding boxes and json predictions when testing: _python3 flow --pbLoad built_graph/yolo-voc-3c.pb --metaLoad built_graph/yolo-voc-3c.meta --imgdir sampleimg/ --json
Does anyone know how to integrate yolo into OpenCV for unity ensuring a working model?
I've trainned my yolo model for 13875 stages as below:
python3 flow --model cfg/yolo-voc-3c.cfg --load bin/tiny-yolo-voc.weights --train --trainer adam --annotation xml/ --dataset output/ --gpu 1.0 --save 2000 --keep 500 --epoch 100 > progress.txt
Then I've saved the result as .pb file, which seems to work on darkflow environment, but it doesn't predict anything when it's integrated into OpenCVForUnity. This integration it's done by setting the dnn net as _Net yoloNet = readNet("pb_pathfile/yolo-voc-3c.pb") , passing the blob input to the net (yoloNet.setinput(blob)) and doing its forward, it returns no boundingbox outputs. But there must being something wrong on that transference since I know the yolo-voc-3c.pb properly works, because it returns both bounding boxes and json predictions when testing: _python3 flow --pbLoad built_graph/yolo-voc-3c.pb --metaLoad built_graph/yolo-voc-3c.meta --imgdir sampleimg/ --json
Does anyone know how to integrate yolo into OpenCV for unity ensuring a working model?