search

smallcorgi / Faster-RCNN_TF

Faster-RCNN in Tensorflow
MIT License
2.34k stars 1.12k forks source link

Visualizing with Tensorboard #280

Open LK1711 opened 6 years ago

LK1711 commented 6 years ago

Has anyone integrated Tensorboard while training/testing their own dataset? I do not understand which file the line of code mentioned here should go:https://www.tensorflow.org/programmers_guide/summaries_and_tensorboard

Could someone please help!

Thanks in advance!

walsvid commented 6 years ago

I have just completed the work of integrating tensorboard into faster-rcnn. The code is base on this repo. For more details you can reference my repo: Visit https://github.com/walsvid/Faster-RCNN-TensorFlow

The added code is mainly summary writer and several scalar summary of loss.

If you want to know the specific implementation details, you can view these commits: walsvid/Faster-RCNN-TensorFlow@82a072a2e07ec1a78eddd08412b0e883584966e0, walsvid/Faster-RCNN-TensorFlow@0f606601995f505e94f80173fb411e09dfe1add5 and walsvid/Faster-RCNN-TensorFlow@7ba482eb836f679d90345b34da1bb70c429ec7d1.

meijie0401 commented 6 years ago

@walsvid Hi thanks for your code for tensorboard. But I also want to see the loss on validation dataset. Do you know how to feed validation dataset to feed_dict ? thanks a lot!!!!!

walsvid commented 6 years ago

@meijie0401 Since the dataset of Faster-rcnn is trainval and test, the roidb used in the current code does not distinguish between train and val, so you can refer to this part of the code to build your model with val_roidb: Reference

meijie0401 commented 6 years ago

@walsvid Thank you!! another quick questoin: is there normalization for input image in this repo? I can't l locate normolization in this code.