GitLab

Internal changes

--

PiperOrigin-RevId: 316837667

* Merged commit includes the following changes:
311933687  by Sergio Guadarrama:

    Removes spurios use of tf.compat.v2, which results in spurious tf.compat.v1.compat.v2. Adds basic test to nasnet_utils.
    Replaces all remaining import tensorflow as tf with import tensorflow.compat.v1 as tf

--
311766063  by Sergio Guadarrama:

    Removes explicit tf.compat.v1 in all call sites (we already import tf.compat.v1, so this code was  doing tf.compat.v1.compat.v1). The existing code worked in latest version of tensorflow, 2.2, (and 1.15) but not in 1.14 or in 2.0.0a, this CL fixes it.

--
311624958  by Sergio Guadarrama:

    Updates README that doesn't render properly in github documentation

--
310980959  by Sergio Guadarrama:

    Moves research_models/slim off tf.contrib.slim/layers/framework to tf_slim

--
310263156  by Sergio Guadarrama:

    Adds model breakdown for MobilenetV3

--
308640...

310447280  by lzc:

    Internal change

310420845  by Zhichao Lu:

    Open source the internal Context RCNN code.

--
310362339  by Zhichao Lu:

    Internal change

310259448  by lzc:

    Update required TF version for OD API.

--
310252159  by Zhichao Lu:

    Port patch_ops_test to TF1/TF2 as TPUs.

--
310247180  by Zhichao Lu:

    Ignore keypoint heatmap loss in the regions/bounding boxes with target keypoint
    class but no valid keypoint annotations.

--
310178294  by Zhichao Lu:

    Opensource MnasFPN
    https://arxiv.org/abs/1912.01106

--
310094222  by lzc:

    Internal changes.

--
310085250  by lzc:

    Internal Change.

--
310016447  by huizhongc:

    Remove unrecognized classes from labeled_classes.

--
310009470  by rathodv:

    Mark batcher.py as TF1 only.

--
310001984  by rathodv:

    Update core/preprocessor.py to be compatible with TF1/TF2..

--
309455035  by Zhichao Lu:

    Makes the freezable_batch_norm_test run w/ v2 behavior.

    The main change is in v2 updates will happen right away when running batchnorm in training mode. So, we need to restore the weights between batchnorm calls to make sure the numerical checks all start from the same place.

--
309425881  by Zhichao Lu:

    Make TF1/TF2 optimizer builder tests explicit.

--
309408646  by Zhichao Lu:

    Make dataset builder tests TF1 and TF2 compatible.

--
309246305  by Zhichao Lu:

    Added the functionality of combining the person keypoints and object detection
    annotations in the binary that converts the COCO raw data to TfRecord.

--
309125076  by Zhichao Lu:

    Convert target_assigner_utils to TF1/TF2.

--
308966359  by huizhongc:

    Support SSD training with partially labeled groundtruth.

--
308937159  by rathodv:

    Update core/target_assigner.py to be compatible with TF1/TF2.

--
308774302  by Zhichao Lu:

    Internal

--
308732860  by rathodv:

    Make core/prefetcher.py  compatible with TF1 only.

--
308726984  by rathodv:

    Update core/multiclass_nms_test.py to be TF1/TF2 compatible.

--
308714718  by rathodv:

    Update core/region_similarity_calculator_test.py to be TF1/TF2 compatible.

--
308707960  by rathodv:

    Update core/minibatch_sampler_test.py to be TF1/TF2 compatible.

--
308700595  by rathodv:

    Update core/losses_test.py to be TF1/TF2 compatible and remove losses_test_v2.py

--
308361472  by rathodv:

    Update core/matcher_test.py to be TF1/TF2 compatible.

--
308335846  by Zhichao Lu:

    Updated the COCO evaluation logics and populated the groundturth area
    information through. This change matches the groundtruth format expected by the
    COCO keypoint evaluation.

--
308256924  by rathodv:

    Update core/keypoints_ops_test.py to be TF1/TF2 compatible.

--
308256826  by rathodv:

    Update class_agnostic_nms_test.py to be TF1/TF2 compatible.

--
308256112  by rathodv:

    Update box_list_ops_test.py to be TF1/TF2 compatible.

--
308159360  by Zhichao Lu:

    Internal change

308145008  by Zhichao Lu:

    Added 'image/class/confidence' field in the TFExample decoder.

--
307651875  by rathodv:

    Refactor core/box_list.py to support TF1/TF2.

--
307651798  by rathodv:

    Modify box_coder.py base class to work with with TF1/TF2

--
307651652  by rathodv:

    Refactor core/balanced_positive_negative_sampler.py to support TF1/TF2.

--
307651571  by rathodv:

    Modify BoxCoders tests to use test_case:execute method to allow testing with TF1.X and TF2.X

--
307651480  by rathodv:

    Modify Matcher tests to use test_case:execute method to allow testing with TF1.X and TF2.X

--
307651409  by rathodv:

    Modify AnchorGenerator tests to use test_case:execute method to allow testing with TF1.X and TF2.X

--
307651314  by rathodv:

    Refactor model_builder to support TF1 or TF2 models based on TensorFlow version.

--
307092053  by Zhichao Lu:

    Use manager to save checkpoint.

--
307071352  by ronnyvotel:

    Fixing keypoint visibilities. Now by default, the visibility is marked True if the keypoint is labeled (regardless of whether it is visible or not).
    Also, if visibilities are not present in the dataset, they will be created based on whether the keypoint coordinates are finite (vis = True) or NaN (vis = False).

--
307069557  by Zhichao Lu:

    Internal change to add few fields related to postprocessing parameters in
    center_net.proto and populate those parameters to the keypoint postprocessing
    functions.

--
307012091  by Zhichao Lu:

    Make Adam Optimizer's epsilon proto configurable.

    Potential issue: tf.compat.v1's AdamOptimizer has a default epsilon on 1e-08 ([doc-link](https://www.tensorflow.org/api_docs/python/tf/compat/v1/train/AdamOptimizer))  whereas tf.keras's AdamOptimizer has default epsilon 1e-07 ([doc-link](https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/Adam))

--
306858598  by Zhichao Lu:

    Internal changes to update the CenterNet model:
    1) Modified eval job loss computation to avoid averaging over batches with zero loss.
    2) Updated CenterNet keypoint heatmap target assigner to apply box size to heatmap Guassian standard deviation.
    3) Updated the CenterNet meta arch keypoint losses computation to apply weights outside of loss function.

--
306731223  by jonathanhuang:

    Internal change.

--
306549183  by rathodv:

    Internal Update.

--
306542930  by rathodv:

    Internal Update

--
306322697  by rathodv:

    Internal.

--
305345036  by Zhichao Lu:

    Adding COCO Camera Traps Json to tf.Example beam code

--
304104869  by lzc:

    Internal changes.

--
304068971  by jonathanhuang:

    Internal change.

--
304050469  by Zhichao Lu:

    Internal change.

--
303880642  by huizhongc:

    Support parsing partially labeled groundtruth.

--
303841743  by Zhichao Lu:

    Deprecate nms_on_host in SSDMetaArch.

--
303803204  by rathodv:

    Internal change.

--
303793895  by jonathanhuang:

    Internal change.

--
303467631  by rathodv:

    Py3 update for detection inference test.

--
303444542  by rathodv:

    Py3 update to metrics module

--
303421960  by rathodv:

    Update json_utils to python3.

--
302787583  by ronnyvotel:

    Coco results generator for submission to the coco test server.

--
302719091  by Zhichao Lu:

    Internal change to add the ResNet50 image feature extractor for CenterNet model.

--
302116230  by Zhichao Lu:

    Added the functions to overlay the heatmaps with images in visualization util
    library.

--
301888316  by Zhichao Lu:

    Fix checkpoint_filepath not defined error.

--
301840312  by ronnyvotel:

    Adding keypoint_scores to visualizations.

--
301683475  by ronnyvotel:

    Introducing the ability to preprocess `keypoint_visibilities`.

    Some data augmentation ops such as random crop can filter instances and keypoints. It's important to also filter keypoint visibilities, so that the groundtruth tensors are always in alignment.

--
301532344  by Zhichao Lu:

    Don't use tf.divide since "Quantization not yet supported for op: DIV"

--
301480348  by ronnyvotel:

    Introducing keypoint evaluation into model lib v2.
    Also, making some fixes to coco keypoint evaluation.

--
301454018  by Zhichao Lu:

    Added the image summary to visualize the train/eval input images and eval's
    prediction/groundtruth side-by-side image.

--
301317527  by Zhichao Lu:

    Updated the random_absolute_pad_image function in the preprocessor library to
    support the keypoints argument.

--
301300324  by Zhichao Lu:

    Apply name change(experimental_run_v2 -> run) for all callers in Tensorflow.

--
301297115  by ronnyvotel:

    Utility function for setting keypoint visibilities based on keypoint coordinates.

--
301248885  by Zhichao Lu:

    Allow MultiworkerMirroredStrategy(MWMS) use by adding checkpoint handling with temporary directories in model_lib_v2. Added missing WeakKeyDictionary cfer_fn_cache field in CollectiveAllReduceStrategyExtended.

--
301224559  by Zhichao Lu:

    ...1) Fixes model_lib to also use keypoints while preparing model groundtruth.
    ...2) Tests model_lib with newly added keypoint metrics config.

--
300836556  by Zhichao Lu:

    Internal changes to add keypoint estimation parameters in CenterNet proto.

--
300795208  by Zhichao Lu:

    Updated the eval_util library to populate the keypoint groundtruth to
    eval_dict.

--
299474766  by Zhichao Lu:

    ...Modifies eval_util to create Keypoint Evaluator objects when configured in eval config.

--
299453920  by Zhichao Lu:

    Add swish activation as a hyperperams option.

--
299240093  by ronnyvotel:

    Keypoint postprocessing for CenterNetMetaArch.

--
299176395  by Zhichao Lu:

    Internal change.

--
299135608  by Zhichao Lu:

    Internal changes to refactor the CenterNet model in preparation for keypoint estimation tasks.

--
298915482  by Zhichao Lu:

    Make dataset_builder aware of input_context for distributed training.

--
298713595  by Zhichao Lu:

    Handling data with negative size boxes.

--
298695964  by Zhichao Lu:

    Expose change_coordinate_frame as a config parameter; fix multiclass_scores optional field.

--
298492150  by Zhichao Lu:

    Rename optimizer_builder_test_v2.py -> optimizer_builder_v2_test.py

--
298476471  by Zhichao Lu:

    Internal changes to support CenterNet keypoint estimation.

--
298365851  by ronnyvotel:

    Fixing a bug where groundtruth_keypoint_weights were being padded with a dynamic dimension.

--
297843700  by Zhichao Lu:

    Internal change.

--
297706988  by lzc:

    Internal change.

--
297705287  by ronnyvotel:

    Creating the "snapping" behavior in CenterNet, where regressed keypoints are refined with updated candidate keypoints from a heatmap.

--
297700447  by Zhichao Lu:

    Improve checkpoint checking logic with TF2 loop.

--
297686094  by Zhichao Lu:

    Convert "import tensorflow as tf" to "import tensorflow.compat.v1".

--
297670468  by lzc:

    Internal change.

--
297241327  by Zhichao Lu:

    Convert "import tensorflow as tf" to "import tensorflow.compat.v1".

--
297205959  by Zhichao Lu:

    Internal changes to support refactored the centernet object detection target assigner into a separate library.

--
297143806  by Zhichao Lu:

    Convert "import tensorflow as tf" to "import tensorflow.compat.v1".

--
297129625  by Zhichao Lu:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
297117070  by Zhichao Lu:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
297030190  by Zhichao Lu:

    Add configuration options for visualizing keypoint edges

--
296359649  by Zhichao Lu:

    Support DepthwiseConv2dNative (of separable conv) in weight equalization loss.

--
296290582  by Zhichao Lu:

    Internal change.

--
296093857  by Zhichao Lu:

    Internal changes to add general target assigner utilities.

--
295975116  by Zhichao Lu:

    Fix visualize_boxes_and_labels_on_image_array to show max_boxes_to_draw correctly.

--
295819711  by Zhichao Lu:

    Adds a flag to visualize_boxes_and_labels_on_image_array to skip the drawing of axis aligned bounding boxes.

--
295811929  by Zhichao Lu:

    Keypoint support in random_square_crop_by_scale.

--
295788458  by rathodv:

    Remove unused checkpoint to reduce repo size on github

--
295787184  by Zhichao Lu:

    Enable visualization of edges between keypoints

--
295763508  by Zhichao Lu:

    [Context RCNN] Add an option to enable / disable cropping feature in the post
    process step in the meta archtecture.

--
295605344  by Zhichao Lu:

    internal change.

--
294926050  by ronnyvotel:

    Adding per-keypoint groundtruth weights. These weights are intended to be used as multipliers in a keypoint loss function.

    Groundtruth keypoint weights are constructed as follows:
    - Initialize the weight for each keypoint type based on user-specified weights in the input_reader proto
    - Mask out (i.e. make zero) all keypoint weights that are not visible.

--
294829061  by lzc:

    Internal change.

--
294566503  by Zhichao Lu:

    Changed internal CenterNet Model configuration.

--
294346662  by ronnyvotel:

    Using NaN values in keypoint coordinates that are not visible.

--
294333339  by Zhichao Lu:

    Change experimetna_distribute_dataset -> experimental_distribute_dataset_from_function

--
293928752  by Zhichao Lu:

    Internal change

--
293909384  by Zhichao Lu:

    Add capabilities to train 1024x1024 CenterNet models.

--
293637554  by ronnyvotel:

    Adding keypoint visibilities to TfExampleDecoder.

--
293501558  by lzc:

    Internal change.

--
293252851  by Zhichao Lu:

    Change tf.gfile.GFile to tf.io.gfile.GFile.

--
292730217  by Zhichao Lu:

    Internal change.

--
292456563  by lzc:

    Internal changes.

--
292355612  by Zhichao Lu:

    Use tf.gather and tf.scatter_nd instead of matrix ops.

--
292245265  by rathodv:

    Internal

--
291989323  by richardmunoz:

    Refactor out building a DataDecoder from building a tf.data.Dataset.

--
291950147  by Zhichao Lu:

    Flip bounding boxes in arbitrary shaped tensors.

--
291401052  by huizhongc:

    Fix multiscale grid anchor generator to allow fully convolutional inference. When exporting model with identity_resizer as image_resizer, there is an incorrect box offset on the detection results. We add the anchor offset to address this problem.

--
291298871  by Zhichao Lu:

    Py3 compatibility changes.

--
290957957  by Zhichao Lu:

    Hourglass feature extractor for CenterNet.

--
290564372  by Zhichao Lu:

    Internal change.

--
290155278  by rathodv:

    Remove Dataset Explorer.

--
290155153  by Zhichao Lu:

    Internal change

--
290122054  by Zhichao Lu:

    Unify the format in the faster_rcnn.proto

--
290116084  by Zhichao Lu:

    Deprecate tensorflow.contrib.

--
290100672  by Zhichao Lu:

    Update MobilenetV3 SSD candidates

--
289926392  by Zhichao Lu:

    Internal change

--
289553440  by Zhichao Lu:

    [Object Detection API] Fix the comments about the dimension of the rpn_box_encodings from 4-D to 3-D.

--
288994128  by lzc:

    Internal changes.

--
288942194  by lzc:

    Internal change.

--
288746124  by Zhichao Lu:

    Configurable channel mean/std. dev in CenterNet feature extractors.

--
288552509  by rathodv:

    Internal.

--
288541285  by rathodv:

    Internal update.

--
288396396  by Zhichao Lu:

    Make object detection import contrib explicitly

--
288255791  by rathodv:

    Internal

--
288078600  by Zhichao Lu:

    Fix model_lib_v2 test

--
287952244  by rathodv:

    Internal

--
287921774  by Zhichao Lu:

    internal change

--
287906173  by Zhichao Lu:

    internal change

--
287889407  by jonathanhuang:

    PY3 compatibility

--
287889042  by rathodv:

    Internal

--
287876178  by Zhichao Lu:

    Internal change.

--
287770490  by Zhichao Lu:

    Add CenterNet proto and builder

--
287694213  by Zhichao Lu:

    Support for running multiple steps per tf.function call.

--
287377183  by jonathanhuang:

    PY3 compatibility

--
287371344  by rathodv:

    Support loading keypoint labels and ids.

--
287368213  by rathodv:

    Add protos supporting keypoint evaluation.

--
286673200  by rathodv:

    dataset_tools PY3 migration

--
286635106  by Zhichao Lu:

    Update code for upcoming tf.contrib removal

--
286479439  by Zhichao Lu:

    Internal change

--
286311711  by Zhichao Lu:

    Skeleton of context model within TFODAPI

--
286005546  by Zhichao Lu:

    Fix Faster-RCNN training when using keep_aspect_ratio_resizer with pad_to_max_dimension

--
285906400  by derekjchow:

    Internal change

--
285822795  by Zhichao Lu:

    Add CenterNet meta arch target assigners.

--
285447238  by Zhichao Lu:

    Internal changes.

--
285016927  by Zhichao Lu:

    Make _dummy_computation a tf.function. This fixes breakage caused by
    cl/284256438

--
284827274  by Zhichao Lu:

    Convert to python 3.

--
284645593  by rathodv:

    Internal change

--
284639893  by rathodv:

    Add missing documentation for keypoints in eval_util.py.

--
284323712  by Zhichao Lu:

    Internal changes.

--
284295290  by Zhichao Lu:

    Updating input config proto and dataset builder to include context fields

    Updating standard_fields and tf_example_decoder to include context features

--
284226821  by derekjchow:

    Update exporter.

--
284211030  by Zhichao Lu:

    API changes in CenterNet informed by the experiments with hourlgass network.

--
284190451  by Zhichao Lu:

    Add support for CenterNet losses in protos and builders.

--
284093961  by lzc:

    Internal changes.

--
284028174  by Zhichao Lu:

    Internal change

--
284014719  by derekjchow:

    Do not pad top_down feature maps unnecessarily.

--
284005765  by Zhichao Lu:

    Add new pad_to_multiple_resizer

--
283858233  by Zhichao Lu:

    Make target assigner work when under tf.function.

--
283836611  by Zhichao Lu:

    Make config getters more general.

--
283808990  by Zhichao Lu:

    Internal change

--
283754588  by Zhichao Lu:

    Internal changes.

--
282460301  by Zhichao Lu:

    Add ability to restore v2 style checkpoints.

--
281605842  by lzc:

    Add option to disable loss computation in OD API eval job.

--
280298212  by Zhichao Lu:

    Add backwards compatible change

--
280237857  by Zhichao Lu:

    internal change

--

PiperOrigin-RevId: 310447280

250447559  by Zhichao Lu:

    Update expected files format for Instance Segmentation challenge:
    - add fields ImageWidth, ImageHeight and store the values per prediction
    - as mask, store only encoded image and assume its size is ImageWidth x ImageHeight

--
250402780  by rathodv:

    Fix failing Mask R-CNN TPU convergence test.

    Cast second stage prediction tensors from bfloat16 to float32 to prevent errors in third target assignment (Mask Prediction) - Concat with different types bfloat16 and bfloat32 isn't allowed.

--
250300240  by Zhichao Lu:

    Addion Open Images Challenge 2019 object detection and instance segmentation
    support into Estimator framework.

--
249944839  by rathodv:

    Modify exporter.py to add multiclass score nodes in exported inference graphs.

--
249935201  by rathodv:

    Modify postprocess methods to preserve multiclass scores after non max suppression.

--
249878079  by Zhichao Lu:

    This CL slightly refactors some Object Detection helper functions for data creation, evaluation, and groundtruth providing.

    This will allow the eager+function custom loops to share code with the existing estimator training loops.

    Concretely we make the following changes:
    1. In input creation we separate dataset-creation into top-level helpers, and allow it to optionally accept a pre-constructed model directly instead of always creating a model from the config just for feature preprocessing.

    2. In coco evaluation we split the update_op creation into its own function, which the custom loops will call directly.

    3. In model_lib we move groundtruth providing/ datastructure munging into a helper function

    4. For now we put an escape hatch in `_summarize_target_assignment` when executing in tf v2.0 behavior because the summary apis used only work w/ tf 1.x

--
249673507  by rathodv:

    Use explicit casts instead of tf.to_float and tf.to_int32 to avoid warnings.

--
249656006  by Zhichao Lu:

    Add named "raw_keypoint_locations" node that corresponds with the "raw_box_locations" node.

--
249651674  by rathodv:

    Keep proposal boxes in float format. MatMulCropAndResize can handle the type even when feature themselves are bfloat16s.

--
249568633  by rathodv:

    Support q > 1 in class agnostic NMS.
    Break post_processing_test.py into 3 separate files to avoid linter errors.

--
249535530  by rathodv:

    Update some deprecated arguments to tf ops.

--
249368223  by rathodv:

    Modify MatMulCropAndResize to use MultiLevelRoIAlign method and move the tests to spatial_transform_ops.py module.

    This cl establishes that CropAndResize and RoIAlign are equivalent and only differ in the sampling point grid within the boxes. CropAndResize uses a uniform size x size point grid such that the corner points exactly overlap box corners, while RoiAlign divides boxes into size x size cells and uses their centers as sampling points. In this cl, we switch MatMulCropAndResize to use the MultiLevelRoIAlign implementation with `align_corner` option as MultiLevelRoIAlign implementation is more memory efficient on TPU when compared to the original MatMulCropAndResize.

--
249337338  by chowdhery:

    Add class-agnostic non-max-suppression in post_processing

--
249139196  by Zhichao Lu:

    Fix positional argument bug in export_tflite_ssd_graph

--
249120219  by Zhichao Lu:

    Add evaluator for computing precision limited to a given recall range.

--
249030593  by Zhichao Lu:

    Evaluation util to run segmentation and detection challenge evaluation.

--
248554358  by Zhichao Lu:

    This change contains the auxiliary changes required for TF 2.0 style training with eager+functions+dist strat loops, but not the loops themselves.

    It includes:
    - Updates to shape usage to support both tensorshape v1 and tensorshape v2
    - A fix to FreezableBatchNorm to not override the `training` arg in call when `None` was passed to the constructor (Not an issue in the estimator loops but it was in the custom loops)
    - Puts some constants in init_scope so they work in eager + functions
    - Makes learning rate schedules return a callable in eager mode (required so they update when the global_step changes)
    - Makes DetectionModel a tf.module so it tracks variables (e.g. ones nested in layers)
    - Removes some references to `op.name` for some losses and replaces it w/ explicit names
    - A small part of the change to allow the coco evaluation metrics to work in eager mode

--
248271226  by rathodv:

    Add MultiLevel RoIAlign op.

--
248229103  by rathodv:

    Add functions to 1. pad features maps 2. ravel 5-D indices

--
248206769  by rathodv:

    Add utilities needed to introduce RoI Align op.

--
248177733  by pengchong:

    Internal changes

--
247742582  by Zhichao Lu:

    Open Images Challenge 2019 instance segmentation metric: part 2

--
247525401  by Zhichao Lu:

    Update comments on max_class_per_detection.

--
247520753  by rathodv:

    Add multilevel crop and resize operation that builds on top of matmul_crop_and_resize.

--
247391600  by Zhichao Lu:

    Open Images Challenge 2019 instance segmentation metric

--
247325813  by chowdhery:

    Quantized MobileNet v2 SSD FPNLite config with depth multiplier 0.75

--

PiperOrigin-RevId: 250447559

Release iNaturalist Species-trained models, refactor of evaluation, box predictor for object detection. (#5289)

* Merged commit includes the following changes:
212389173  by Zhichao Lu:

    1. Replace tf.boolean_mask with tf.where

--
212282646  by Zhichao Lu:

    1. Fix a typo in model_builder.py and add a test to cover it.

--
212142989  by Zhichao Lu:

    Only resize masks in meta architecture if it has not already been resized in the input pipeline.

--
212136935  by Zhichao Lu:

    Choose matmul or native crop_and_resize in the model builder instead of faster r-cnn meta architecture.

--
211907984  by Zhichao Lu:

    Make eval input reader repeated field and update config util to handle this field.

--
211858098  by Zhichao Lu:

    Change the implementation of merge_boxes_with_multiple_labels.

--
211843915  by Zhichao Lu:

    Add Mobilenet v2 + FPN support.

--
211655076  by Zhichao Lu:

    Bug fix for generic keys in config overrides

    In generic configuration overrides, we had a duplicate entry for train_input_config and we were missing the eval_input_config and eval_config.

    This change also introduces testing for all config overrides.

--
211157501  by Zhichao Lu:

    Make the locally-modified conv defs a copy.

    So that it doesn't modify MobileNet conv defs globally for other code that
    transitively imports this package.

--
211112813  by Zhichao Lu:

    Refactoring visualization tools for Estimator's eval_metric_ops. This will make it easier for future models to take advantage of a single interface and mechanics.

--
211109571  by Zhichao Lu:

    A test decorator.

--
210747685  by Zhichao Lu:

    For FPN, when use_depthwise is set to true, use slightly modified mobilenet v1 config.

--
210723882  by Zhichao Lu:

    Integrating the losses mask into the meta architectures. When providing groundtruth, one can optionally specify annotation information (i.e. which images are labeled vs. unlabeled). For any image that is unlabeled, there is no loss accumulation.

--
210673675  by Zhichao Lu:

    Internal change.

--
210546590  by Zhichao Lu:

    Internal change.

--
210529752  by Zhichao Lu:

    Support batched inputs with ops.matmul_crop_and_resize.

    With this change the new inputs are images of shape [batch, heigh, width, depth] and boxes of shape [batch, num_boxes, 4]. The output tensor is of the shape [batch, num_boxes, crop_height, crop_width, depth].

--
210485912  by Zhichao Lu:

    Fix TensorFlow version check in object_detection_tutorial.ipynb

--
210484076  by Zhichao Lu:

    Reduce TPU memory required for single image matmul_crop_and_resize.

    Using tf.einsum eliminates intermediate tensors, tiling and expansion. for an image of size [40, 40, 1024] and boxes of shape [300, 4] HBM memory usage goes down from 3.52G to 1.67G.

--
210468361  by Zhichao Lu:

    Remove PositiveAnchorLossCDF/NegativeAnchorLossCDF to resolve "Main thread is not in main loop error" issue in local training.

--
210100253  by Zhichao Lu:

    Pooling pyramid feature maps: add option to replace max pool with convolution layers.

--
209995842  by Zhichao Lu:

    Fix a bug which prevents variable sharing in Faster RCNN.

--
209965526  by Zhichao Lu:

    Add support for enabling export_to_tpu through the estimator.

--
209946440  by Zhichao Lu:

    Replace deprecated tf.train.Supervisor with tf.train.MonitoredSession. MonitoredSession also takes away the hassle of starting queue runners.

--
209888003  by Zhichao Lu:

    Implement function to handle data where source_id is not set.

    If the field source_id is found to be the empty string for any image during runtime, it will be replaced with a random string. This avoids hash-collisions on dataset where many examples do not have source_id set. Those hash-collisions have unintended site effects and may lead to bugs in the detection pipeline.

--
209842134  by Zhichao Lu:

    Converting loss mask into multiplier, rather than using it as a boolean mask (which changes tensor shape). This is necessary, since other utilities (e.g. hard example miner) require a loss matrix with the same dimensions as the original prediction tensor.

--
209768066  by Zhichao Lu:

    Adding ability to remove loss computation from specific images in a batch, via an optional boolean mask.

--
209722556  by Zhichao Lu:

    Remove dead code.

    (_USE_C_API was flipped to True by default in TensorFlow 1.8)

--
209701861  by Zhichao Lu:

    This CL cleans-up some tf.Example creation snippets, by reusing the convenient tf.train.Feature building functions in dataset_util.

--
209697893  by Zhichao Lu:

    Do not overwrite num_epoch for eval input. This leads to errors in some cases.

--
209694652  by Zhichao Lu:

    Sample boxes by jittering around the currently given boxes.

--
209550300  by Zhichao Lu:

    `create_category_index_from_labelmap()` function now accepts `use_display_name` parameter.
    Also added create_categories_from_labelmap function for convenience

--
209490273  by Zhichao Lu:

    Check result_dict type before accessing image_id via key.

--
209442529  by Zhichao Lu:

    Introducing the capability to sample examples for evaluation. This makes it easy to specify one full epoch of evaluation, or a subset (e.g. sample 1 of every N examples).

--
208941150  by Zhichao Lu:

    Adding the capability of exporting the results in json format.

--
208888798  by Zhichao Lu:

    Fixes wrong dictionary key for num_det_boxes_per_image.

--
208873549  by Zhichao Lu:

    Reduce the number of HLO ops created by matmul_crop_and_resize.

    Do not unroll along the channels dimension. Instead, transpose the input image dimensions, apply tf.matmul and transpose back.

    The number of HLO instructions for 1024 channels reduce from 12368 to 110.

--
208844315  by Zhichao Lu:

    Add an option to use tf.non_maximal_supression_padded in SSD post-process

--
208731380  by Zhichao Lu:

    Add field in box_predictor config to enable mask prediction and update builders accordingly.

--
208699405  by Zhichao Lu:

    This CL creates a keras-based multi-resolution feature map extractor.

--
208557208  by Zhichao Lu:

    Add TPU tests for Faster R-CNN Meta arch.

    * Tests that two_stage_predict and total_loss tests run successfully on TPU.
    * Small mods to multiclass_non_max_suppression to preserve static shapes.

--
208499278  by Zhichao Lu:

    This CL makes sure the Keras convolutional box predictor & head layers apply activation layers *after* normalization (as opposed to before).

--
208391694  by Zhichao Lu:

    Updating visualization tool to produce multiple evaluation images.

--
208275961  by Zhichao Lu:

    This CL adds a Keras version of the Convolutional Box Predictor, as well as more general infrastructure for making Keras Prediction heads & Keras box predictors.

--
208275585  by Zhichao Lu:

    This CL enables the Keras layer hyperparameter object to build a dedicated activation layer, and to disable activation by default in the op layer construction kwargs.

    This is necessary because in most cases the normalization layer must be applied before the activation layer. So, in Keras models we must set the convolution activation in a dedicated layer after normalization is applied, rather than setting it in the convolution layer construction args.

--
208263792  by Zhichao Lu:

    Add a new SSD mask meta arch that can predict masks for SSD models.
    Changes including:
     - overwrite loss function to add mask loss computation.
     - update ssd_meta_arch to handle masks if predicted in predict and postprocessing.

--
208000218  by Zhichao Lu:

    Make FasterRCNN choose static shape operations only in training mode.

--
207997797  by Zhichao Lu:

    Add static boolean_mask op to box_list_ops.py and use that in faster_rcnn_meta_arch.py to support use_static_shapes option.

--
207993460  by Zhichao Lu:

    Include FGVC detection models in model zoo.

--
207971213  by Zhichao Lu:

    remove the restriction to run tf.nn.top_k op on CPU

--
207961187  by Zhichao Lu:

    Build the first stage NMS function in the model builder and pass it to FasterRCNN meta arch.

--
207960608  by Zhichao Lu:

    Internal Change.

--
207927015  by Zhichao Lu:

    Have an option to use the TPU compatible NMS op cl/206673787, in the batch_multiclass_non_max_suppression function. On setting pad_to_max_output_size to true, the output nmsed boxes are padded to be of length max_size_per_class.

    This can be used in first stage Region Proposal Network in FasterRCNN model by setting the first_stage_nms_pad_to_max_proposals field to true in config proto.

--
207809668  by Zhichao Lu:

    Add option to use depthwise separable conv instead of conv2d in FPN and WeightSharedBoxPredictor. More specifically, there are two related configs:
    - SsdFeatureExtractor.use_depthwise
    - WeightSharedConvolutionalBoxPredictor.use_depthwise

--
207808651  by Zhichao Lu:

    Fix the static balanced positive negative sampler's TPU tests

--
207798658  by Zhichao Lu:

    Fixes a post-refactoring bug where the pre-prediction convolution layers in the convolutional box predictor are ignored.

--
207796470  by Zhichao Lu:

    Make slim endpoints visible in FasterRCNNMetaArch.

--
207787053  by Zhichao Lu:

    Refactor ssd_meta_arch so that the target assigner instance is passed into the SSDMetaArch constructor rather than constructed inside.

--

PiperOrigin-RevId: 212389173

* Fix detection model zoo typo.

* Modify tf example decoder to handle label maps with either `display_name` or `name` fields seamlessly.

Currently, tf example decoder uses only `name` field to look up ids for class text field present in the data. This change uses both `display_name` and `name` fields in the label map to fetch ids for class text.

PiperOrigin-RevId: 212672223

* Modify create_coco_tf_record tool to write out class text instead of class labels.

PiperOrigin-RevId: 212679112

* Fix detection model zoo typo.

PiperOrigin-RevId: 212715692

* Adding the following two optional flags to WeightSharedConvolutionalBoxHead:
1) In the box head, apply clipping to box encodings in the box head.
2) In the class head, apply sigmoid to class predictions at inference time.

PiperOrigin-RevId: 212723242

* Support class confidences in merge boxes with multiple labels.

PiperOrigin-RevId: 212884998

* Creates multiple eval specs for object detection.

PiperOrigin-RevId: 212894556

* Set batch_norm on last layer in Mask Head to None.

PiperOrigin-RevId: 213030087

* Enable bfloat16 training for object detection models.

PiperOrigin-RevId: 213053547

* Skip padding op when unnecessary.

PiperOrigin-RevId: 213065869

* Modify `Matchers` to use groundtruth weights before performing matching.

Groundtruth weights tensor is used to indicate padding in groundtruth box tensor. It is handled in `TargetAssigner` by creating appropriate classification and regression target weights based on the groundtruth box each anchor matches to. However, options such as `force_match_all_rows` in `ArgmaxMatcher` force certain anchors to match to groundtruth boxes that are just paddings thereby reducing the number of anchors that could otherwise match to real groundtruth boxes.

For single stage models like SSD the effect of this is negligible as there are two orders of magnitude more anchors than the number of padded groundtruth boxes. But for Faster R-CNN and Mask R-CNN where there are only 300 anchors in the second stage, a significant number of these match to groundtruth paddings reducing the number of anchors regressing to real groundtruth boxes degrading the performance severely.

Therefore, this change introduces an additional boolean argument `valid_rows` to `Matcher.match` methods and the implementations now ignore such padded groudtruth boxes during matching.

PiperOrigin-RevId: 213345395

* Add release note for iNaturalist Species trained models.

PiperOrigin-RevId: 213347179

* Fix the bug of uninitialized gt_is_crowd_list variable.

PiperOrigin-RevId: 213364858

* ...text exposed to open source public git repo...

PiperOrigin-RevId: 213554260

187187978  by Zhichao Lu:

    Only updating hyperparameters if they have non-null values.

--
187097690  by Zhichao Lu:

    Rewrite some conditions a bit more clearly.

--
187085190  by Zhichao Lu:

    More informative error message.

--
186935376  by Zhichao Lu:

    Added option to evaluator.evaluate to use custom evaluator objects.

--
186808249  by Zhichao Lu:

    Fix documentation re: number of stages.

--
186775014  by Zhichao Lu:

    Change anchor generator interface to return a list of BoxLists containing anchors for different feature map layers.

--
186729028  by Zhichao Lu:

    Minor fixes to object detection.

--
186723716  by Zhichao Lu:

    Fix tf_example_decoder.py initailization issue.

--
186668505  by Zhichao Lu:

    Remove unused import.

--
186475361  by Zhichao Lu:

    Update the box predictor interface to return list of predictions - one from each feature map - instead of stacking them into one large tensor.

--
186410844  by Zhichao Lu:

    Fix PythonPath Dependencies.

--
186365384  by Zhichao Lu:

    Made some of the functions in exporter public so they can be reused.

--
186341438  by Zhichao Lu:

    Re-introducing check that label-map-path must be a valid (non-empty) string prior to overwriting pipeline config.

--
186036984  by Zhichao Lu:

    Adding default hyperparameters and allowing for overriding them via flags.

--
186026006  by Zhichao Lu:

    Strip `eval_` prefix from name argument give to TPUEstimator.evaluate since it adds the same prefix internally.

--
186016042  by Zhichao Lu:

    Add an option to evaluate models on training data.

--
185944986  by Zhichao Lu:

    let _update_label_map_path go through even if the path is empty

--
185860781  by Zhichao Lu:

    Add random normal initializer option to hyperparams builder.

    Scale the regression losses outside of the box encoder by adjusting huber loss delta and regression loss weight.

--
185846325  by Zhichao Lu:

    Add an option to normalize localization loss by the code size(number of box coordinates) in SSD Meta architecture.

--
185761217  by Zhichao Lu:

    Change multiscale_grid_anchor_generator to return anchors in normalized coordinates by default and add option to configure it.

    In SSD meta architecture, TargetAssigner operates in normalized coordinate space (i.e, groundtruth boxes are in normalized coordinates) hence we need the option to generate anchors in normalized coordinates.

--
185747733  by Zhichao Lu:

    Change the smooth L1 localization implementationt to use tf.losses.huber_loss and expose the delta parameter in the proto.

--
185715309  by Zhichao Lu:

    Obviates the need for prepadding on mobilenet v1 and v2 for fully convolutional models.

--
185685695  by Zhichao Lu:

    Fix manual stepping schedule to return first rate when there are no boundaries

--
185621650  by Zhichao Lu:

    Added target assigner proto for configuring negative class weights.

--

PiperOrigin-RevId: 187187978

185215255  by Zhichao Lu:

    Stop populating image/object/class/text field when generating COCO tf record.

--
185213306  by Zhichao Lu:

    Use the params batch size and not the one from train_config in input_fn

--
185209081  by Zhichao Lu:

    Handle the case when there are no ground-truth masks for an image.

--
185195531  by Zhichao Lu:

    Remove unstack and stack operations on features from third_party/object_detection/model.py.

--
185195017  by Zhichao Lu:

    Matrix multiplication based gather op implementation.

--
185187744  by Zhichao Lu:

    Fix eval_util minor issue.

--
185098733  by Zhichao Lu:

    Internal change

185076656  by Zhichao Lu:

    Increment the amount of boxes for coco17.

--
185074199  by Zhichao Lu:

    Add config for SSD Resnet50 v1 with FPN.

--
185060199  by Zhichao Lu:

    Fix a bug in clear_detections.
    This method set detection_keys to an empty dictionary instead of an empty set. I've refactored so that this ...

184048729  by Zhichao Lu:

    Modify target_assigner so that it creates regression targets taking keypoints into account.

--
184027183  by Zhichao Lu:

    Resnet V1 FPN based feature extractors for SSD meta architecture in Object Detection V2 API.

--
184004730  by Zhichao Lu:

    Expose a lever to override the configured mask_type.

--
183933113  by Zhichao Lu:

    Weight shared convolutional box predictor as described in https://arxiv.org/abs/1708.02002

--
183929669  by Zhichao Lu:

    Expanding box list operations for future data augmentations.

--
183916792  by Zhichao Lu:

    Fix unrecognized assertion function in tests.

--
183906851  by Zhichao Lu:

    - Change ssd meta architecture to use regression weights to compute loss normalizer.

--
183871003  by Zhichao Lu:

    Fix config_util_test wrong dependency.

--
183782120  by Zhichao Lu:

    Add __init__ file to third_party directories.

--
183779109  by Zhichao Lu:

    Setup regular version s...