master
Move the research models into a research subfolder (#2430)
Release MobileDet code and model, and require tf_slim installation for OD API. (#8562)
Add option to use tf.compat.v1.image.resize_nearest_neighbor (instead of PACK and RESHAPE) for upsampling in efficientdet bifpn.
Remove unused comments related to Python 2 compatibility.
Updated the config_util to support the hyperparameter search over new parameters in the CenterNet model.
cleanup of deprecated test methods
Object detection changes: (#7208)
Adding float feature to dataset_util
Open source MnasFPN and minor fixes to OD API (#8484)
Support using the default label map validator by specifying None value.
Added the get_keypoint_label_map_dict function to prepare the merged dictionary that maps from the keypoint names to their ids.
Add an exponential decay learning rate schedule with warmup.
[numpy] Fix users of NumPy APIs that are removed in NumPy 2.0.
Merged commit includes the following changes: (#8740)
Refactor tests for Object Detection API. (#8688)
Fix typo in iou op docstring.
Add partial label evaluation to PrecisionAtRecallEvaluation.
Avoids generating 6D tensors in `nearest_neighbor_upsampling`.
Extend test checking logic to match tpu tfrt's behavior.
Enables combined static and dynamic shape for the batch size in 'MultiLevelRoIAlign'
Merge pull request #8746 from syiming:add_multilevel_crop_and_resize
Add IOU heatmap loss for CenterNet.
Add checkpoint type 'full' to SSD meta arch.
Standardize fine tune checkpoints across all TF2 models.
Update code to be compatible with Pillow 10.
[NumPy] Remove references to deprecated NumPy type aliases.