Hi great work with this project, thanks! I was able to train a model with my own dataset and your training.py. Inspired by this repo, I'm trying to convert the model to TF Lite it is slightly different from the original Yolov3:
In the model I produced using training.py I observe a section around the UpSampling2D op which does not seem to be in the coco_tiny_v3 model displayed on the right. This UpSampling2D operation cannot be used in quantized models so I'm wondering if there is a way to get rid of it (as it does not seem to be part of the original model either)?
It would be awesome if I could convert my yolov3 tiny model to Coral/EdgeTPU!
Hi great work with this project, thanks! I was able to train a model with my own dataset and your training.py. Inspired by this repo, I'm trying to convert the model to TF Lite it is slightly different from the original Yolov3:
In the model I produced using training.py I observe a section around the
UpSampling2Dop which does not seem to be in the coco_tiny_v3 model displayed on the right. ThisUpSampling2Doperation cannot be used in quantized models so I'm wondering if there is a way to get rid of it (as it does not seem to be part of the original model either)?It would be awesome if I could convert my yolov3 tiny model to Coral/EdgeTPU!