I wanted to know is it possible to get per image loss as it helps in finding failure cases for the model and also can we see tensorboard logs ?
Which network will you train, classification or detection(detectnet_v2,yolo,ssd,faster-rcnn,dssd,retinanet)?
For frcnn, currently users can only run tlt-evaluate against the tlt model to check the loss(or mAP) info.
Next release, it will implement evaluation directly during training, similar to detectnet_v2, yolo, ssd networks.
Hello @Morganh, will the next release also implement tensorboard event logging for
No, it is not supported for next release.