Migrate to PyTorch #117
Merged
Migrate to PyTorch #117
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The linear AE is faster than the convolutional one and produces almost the same results. One big issue was the torch.relu(x) on the AE output. This completely wrecked the results. Now it's almost indistinguishable from the TensorFlow implementation.
MMDet expects bboxes as Numpy datatype which is only possible if the annotations are stored as pickle file.
This produced a much better detector performance, possibly because of more negative examples and objects not always being in the image canter.
JpegCompression is deprecated.
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This implements the novelty detection in PyTorch and the object detection with MMDetection.
Resolves #96
Resolves #64
References #91 and #92 as the novelty detection now works with large images.
Maybe #28 and #73 cen be resolved here, too.
Upgrade information:
The
MAIA_AVAILABLE_BYTESenv variable was removed. Instead, the object detection training batch size can be set directly withMAIA_MMDET_TRAIN_BATCH_SIZE.The
COCO_MODEL_URLenv variable was removed. UseMAIA_BACKBONE_MODEL_URLfor the PyTorch backbone file andMAIA_MODEL_URLfir the MMDetection checkpoint file.The training scheme is no longer configurable. Instead of iteration-based training (which still was called "epochs" before), training runs for a fixed number of 12 epochs now (which is the default training config for many MMdet object detectors and worked well in tests).
What's previously been called "instance segmentation" is now correctly called "object detection".