Commits · 09bc9f54fb7084b7908447572938b2e203d7c232 · ModelZoo / ResNet50_tensorflow

03 Dec, 2021 1 commit

Unfreeze keras version as incorrect version issue was fixed. This also... · 59aa6732

Khanh LeViet authored Dec 02, 2021

Unfreeze keras version as incorrect version issue was fixed. This also resolved the endless TF ODT API pip install issue.

PiperOrigin-RevId: 413804051

59aa6732

23 Nov, 2021 1 commit
- Fix conditional convs by adding ReLU. · 65407126
  Vighnesh Birodkar authored Nov 23, 2021
```
PiperOrigin-RevId: 411887283
```
  65407126
19 Nov, 2021 1 commit
- Add multi-layer conditional mask prediction head. · eea839c4
  Vighnesh Birodkar authored Nov 18, 2021
```
PiperOrigin-RevId: 410930769
```
  eea839c4
17 Nov, 2021 1 commit

Generalize the xml parsing to facilitate custom datasets. · 743c28aa

Scott Main authored Nov 17, 2021

Instead of searching the annotation files based on class names in the filenames, read all the XML annotation files. For the cats/dogs use-case, just delete all the XML files that are not the breeds we care about.

PiperOrigin-RevId: 410613750

743c28aa

16 Nov, 2021 1 commit
- Added README.md (#10259) · e0a16ee8
  Jekaterina Jaroslavceva authored Nov 17, 2021
```
* Added README.md

* fix README

* review fixes
```
  e0a16ee8
04 Nov, 2021 2 commits
- Updated the postprocessing logic to perform the keypoint rescoring after NMS · 17169b3b
  Yu-hui Chen authored Nov 04, 2021
```
such that NMS won't wash out the rescored instance scores.

PiperOrigin-RevId: 407602744
```
  17169b3b
- Unify computation of weak losses and fix color consistency computation. · 88d844e7
  Vighnesh Birodkar authored Nov 04, 2021
```
Also upgrades keras version due to
https://github.com/tensorflow/tensorflow/issues/51592

PiperOrigin-RevId: 407556344
```
  88d844e7
03 Nov, 2021 1 commit
- Add link to notebook that generates anchor box ratios with k-means clustering · 6cf9a28c
  Scott Main authored Nov 03, 2021
```
PiperOrigin-RevId: 407366384
```
  6cf9a28c
01 Nov, 2021 2 commits
- Implement the color consistency loss in DeepMAC. · b9dd8e7c
  Vighnesh Birodkar authored Nov 01, 2021
```
PiperOrigin-RevId: 406939681
```
  b9dd8e7c
- List WeightedDiceClassificationLoss at top of docstring. · b1f77678
  Vighnesh Birodkar authored Nov 01, 2021
```
PiperOrigin-RevId: 406929565
```
  b1f77678
25 Oct, 2021 1 commit
- Attempt to fix gh issue #10330 · 12bbefce
  Vighnesh Birodkar authored Oct 25, 2021
```
PiperOrigin-RevId: 405458782
```
  12bbefce
19 Oct, 2021 2 commits
- Updated the tf_example_decoder to support the case where the keypoint labels in · bd303e0a
  Yu-hui Chen authored Oct 19, 2021
```
the image/object/keypoint/text feature but not in the keypoint label map.

PiperOrigin-RevId: 404372105
```
  bd303e0a
- Update keypoint estimation to always use a safe k in top_k · dd933006
  A. Unique TensorFlower authored Oct 19, 2021
```
PiperOrigin-RevId: 404311441
```
  dd933006
13 Oct, 2021 2 commits

Add SpaghettiNet Feature Extractor · ac8d0651
A. Unique TensorFlower authored Oct 13, 2021
```
PiperOrigin-RevId: 402944074
```
ac8d0651

- Added configurable arguments in CenterNet for maximum normalized coordinate... · 218ad3be

A. Unique TensorFlower authored Oct 13, 2021

- Added configurable arguments in CenterNet for maximum normalized coordinate (this might be useful for handling cases where the bounding box can only be presented out-of-bound with the direction-of-path point at the center in panorama rotation).

PiperOrigin-RevId: 402882653

218ad3be

12 Oct, 2021 1 commit
- Clean up the comments that was pointing to the legacy Keras code base. · 66e25b31
  Scott Zhu authored Oct 12, 2021
```
PiperOrigin-RevId: 402663280
```
  66e25b31
05 Oct, 2021 1 commit
- Fix broken link · 75b016b4
  Anirudh Sriram authored Oct 05, 2021
```
PiperOrigin-RevId: 401003754
```
  75b016b4
01 Oct, 2021 2 commits
- Modify CenterNet to output extracted features. · d5ae12e1
  A. Unique TensorFlower authored Sep 30, 2021
```
PiperOrigin-RevId: 400090738
```
  d5ae12e1
- Updated the tf_example_decoder such that it supports combining multiple · e06d2c3a
  Yu-hui Chen authored Sep 30, 2021
```
datasets which contain different subset of keypoints.

PiperOrigin-RevId: 400065449
```
  e06d2c3a
30 Sep, 2021 1 commit
- Added the get_keypoint_label_map_dict function to prepare the merged dictionary · b9c61118
  Yu-hui Chen authored Sep 30, 2021
```
that maps from the keypoint names to their ids.

PiperOrigin-RevId: 400009398
```
  b9c61118
27 Sep, 2021 2 commits
- Ensure that CenterNetMetaArch can process keypoints even when no instances are present. · 87769dc6
  A. Unique TensorFlower authored Sep 27, 2021
```
PiperOrigin-RevId: 399303089
```
  87769dc6
- Limit k of top_k to the size of the dynamic input. · fcb152bf
  A. Unique TensorFlower authored Sep 27, 2021
```
PiperOrigin-RevId: 399210729
```
  fcb152bf
23 Sep, 2021 1 commit
- Make implementation reframe_box_masks_to_image_masks to be more clear. · 58644b96
  Vighnesh Birodkar authored Sep 23, 2021
```
PiperOrigin-RevId: 398596089
```
  58644b96
22 Sep, 2021 1 commit
- Updated the MoveNet.Multipose model postprocessing logics such that it avoids · b1d0bf77
  Yu-hui Chen authored Sep 22, 2021
```
using 5d tensors to avoid issue when converted to TfLite model.

PiperOrigin-RevId: 398301856
```
  b1d0bf77
17 Sep, 2021 2 commits
- INTERNAL CHANGE. · 4b617781
  A. Unique TensorFlower authored Sep 17, 2021
```
PiperOrigin-RevId: 397358689
```
  4b617781
- Check in seq_flow_lite (#10262) · d31fafbf
  pyoung2778 authored Sep 16, 2021
  
  d31fafbf
14 Sep, 2021 1 commit
- Checkin seq_flow_lite (#10219) · fec0338f
  pyoung2778 authored Sep 14, 2021
  
  fec0338f
09 Sep, 2021 1 commit
- Fixes invalid json (#10250) · 0d9ef8d3
  Mark Sandler authored Sep 09, 2021
```
* Fixes invalid json

* Fixed invalid javascript

* fixed json
```
  0d9ef8d3
08 Sep, 2021 1 commit
- Explicit signatures for tflite. Using ideas from #9688 (#10248) · b66b0b05
  Dan Ellis authored Sep 08, 2021
  
  b66b0b05
07 Sep, 2021 1 commit
- Switch out euclidean distance to projection based distance and 1D ResNet. · 9c28cff8
  Vighnesh Birodkar authored Sep 07, 2021
```
PiperOrigin-RevId: 395297294
```
  9c28cff8
02 Sep, 2021 4 commits
- Add example Colab to generate dataset specific aspect ratios using Kmeans · 2cd6f1c6
  A. Unique TensorFlower authored Sep 02, 2021
```
PiperOrigin-RevId: 394568069
```
  2cd6f1c6
- Implement embedding based similarity mask head for DeepMAC. · 3324d816
  Vighnesh Birodkar authored Sep 02, 2021
```
PiperOrigin-RevId: 394512536
```
  3324d816
- change on track_reid classification net to solve dimension error in centerMOT pipeline · 0daae829
  A. Unique TensorFlower authored Sep 02, 2021
```
PiperOrigin-RevId: 394505975
```
  0daae829
- Fixed deprecated tf.contrib.metrics (#9916) · 27bd23e5
  Leandro Schelb authored Sep 02, 2021
```
* Fixed deprecated tf.flags

* Fixed deprecated tf.contrib.metrics
```
  27bd23e5
01 Sep, 2021 3 commits
- Update readme and add link to deeplab2 (#10233) · 0e2d1f0f
  Yukun Zhu authored Sep 01, 2021
  
  0e2d1f0f
- Rewrote the postprocessing logics of the multi-class keypoint task in the · 6dd70405
  Yu-hui Chen authored Aug 31, 2021
```
CenterNet meta arch to avoid using scatter_nd and tf.cond.

Also, use static_or_dynamic_map_fn in the convert_strided_predictions_to_normalized_keypoints function to avoid introducing cycles in the graph that causes the TPU converter errors.

PiperOrigin-RevId: 394152521
```
  6dd70405
- New centernet multi-pose postprocessing logics that avoid using topk op which · 163ca152
  Yu-hui Chen authored Aug 31, 2021
```
runs much slower in the browser.

PiperOrigin-RevId: 394148346
```
  163ca152
28 Aug, 2021 1 commit
- Make ClipWindow with Combined work with batch size > 1. · cd700eaf
  Vivek Rathod authored Aug 27, 2021
```
PiperOrigin-RevId: 393480597
```
  cd700eaf
24 Aug, 2021 1 commit

Fix the order of the input tensors · 843111ca

Eric Forkosh authored Aug 24, 2021

The order of the input tensors is incorrect and therefore the sample notebook does not run.

The sample code has the order of the input tensors as `boxes, classes, scores, num_detections`, but actually the order should be `scores, boxes, num_detections, classes`

After correcting the indexes of the input tensor, the sample notebook runs.

843111ca

20 Aug, 2021 1 commit
- Support clip_window option with Combined NMS. · e31d1693
  Vivek Rathod authored Aug 20, 2021
```
PiperOrigin-RevId: 392061572
```
  e31d1693