open-mmlab · GT9505 · Sep 24, 2021 · Jul 26, 2021 · Jul 26, 2021 · Jul 27, 2021
diff --git a/configs/sot/siamese_rpn/README.md b/configs/sot/siamese_rpn/README.md
@@ -28,18 +28,33 @@ We observe around 1.0 points fluctuations in Success and 1.5 points fluctuations
 
 ### UAV123
 
-The checkpoints from 10-th to 20-th epoch will be evaluated during training.
+The checkpoints from 10-th to 20-th epoch will be evaluated during training. You can find the best checkpoint from the log file.
 
-After training, you need to pick up the best checkpoint from the log file, then use the best checkpoint to search the hyperparameters on UAV123 following [here](https://github.com/open-mmlab/mmtracking/blob/master/docs/useful_tools_scripts.md#siameserpn-test-time-parameter-search) to achieve the best results.
+If you want to get better results, you can use the best checkpoint to search the hyperparameters on UAV123 following [here](https://github.com/open-mmlab/mmtracking/blob/master/docs/useful_tools_scripts.md#siameserpn-test-time-parameter-search).
+Experimentally, the hyperparameters search on UAV123 can bring around 1.0 Success gain.
 
-|    Backbone     |  Style  | Lr schd | Mem (GB) | Inf time (fps) | Success | Norm precision | Config | Download |
-| :-------------: | :-----: | :-----: | :------: | :------------: | :----: | :----: | :------: | :--------: |
-|    R-50    |  -  |   20e    | 7.54     | -              | 61.8 | 77.3 | [config](siamese_rpn_r50_1x_uav123.py) | [model](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_lasot/siamese_rpn_r50_1x_lasot_20201218_051019-3c522eff.pth) &#124; [log](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_lasot/siamese_rpn_r50_1x_lasot_20201218_051019.log.json) |
+The results below are achieved without hyperparameters search. We observe less than 0.5 points fluctuations both in Success and Percision.
+
+|    Backbone     |  Style  | Lr schd | Mem (GB) | Inf time (fps) | Success | Norm Precision | Precision | Config | Download |
+| :-------------: | :-----: | :-----: | :------: | :------------: | :----: | :----: | :------: | :------: | :--------: |
+|    R-50    |  -  |   20e    | 7.54     | -             | 60.6 | 76.5 | 80.5 | [config](siamese_rpn_r50_1x_uav123.py) | [model](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_uav123/siamese_rpn_r50_1x_uav123_20210917_104452-36ac4934.pth) &#124; [log](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_uav123/siamese_rpn_r50_1x_uav123_20210917_104452.log.json) |
 
 ### TrackingNet
 
-The best model on LaSOT is submitted to [the evaluation server on TrackingNet Chanllenge](http://eval.tracking-net.org/web/challenges/challenge-page/39/submission). We provide the best model with its configuration and training log.
+The results of SiameseRPN++ in TrackingNet are reimplemented by ourselves. The best model on LaSOT is submitted to [the evaluation server on TrackingNet Chanllenge](http://eval.tracking-net.org/web/challenges/challenge-page/39/submission). We observe less than 0.5 points fluctuations both in Success and Percision. We provide the best model with its configuration and training log.
 
-|    Backbone     |  Style  | Lr schd | Mem (GB) | Inf time (fps) | Success | Norm precision | Config | Download |
-| :-------------: | :-----: | :-----: | :------: | :------------: | :----: | :----: | :------: | :--------: |
-|    R-50    |  -  |   20e    |      | -              | 70.6 | 77.6 | [config](siamese_rpn_r50_1x_trackingnet.py) | [model](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_lasot/siamese_rpn_r50_1x_lasot_20201218_051019-3c522eff.pth) &#124; [log](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_lasot/siamese_rpn_r50_1x_lasot_20201218_051019.log.json) |
+|    Backbone     |  Style  | Lr schd | Mem (GB) | Inf time (fps) | Success | Norm precision | Precision |Config | Download |
+| :-------------: | :-----: | :-----: | :------: | :------------: | :----: | :----: | :------: | :------: | :--------: |
+|    R-50    |  -  |   20e    |  7.54     | -             | 70.6 | 77.6 | 65.7 | [config](siamese_rpn_r50_1x_trackingnet.py) | [model](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_lasot/siamese_rpn_r50_1x_lasot_20201218_051019-3c522eff.pth) &#124; [log](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_lasot/siamese_rpn_r50_1x_lasot_20201218_051019.log.json) |
+
+### OTB100
+
+The checkpoints from 10-th to 20-th epoch will be evaluated during training. You can find the best checkpoint from the log file.
+
+If you want to get better results, you can use the best checkpoint to search the hyperparameters on OTB100 following [here](https://github.com/open-mmlab/mmtracking/blob/master/docs/useful_tools_scripts.md#siameserpn-test-time-parameter-search). Experimentally, the hyperparameters search on OTB100 can bring around 1.0 Success gain.
+
+**Note:** We train the SiameseRPN++ in the official [pysot](https://github.com/STVIR/pysot) codebase and can not reproduce the same results reported in the paper. We only get 66.1 Success and 86.7 Precision by following the training and hyperparameters searching instructions of pysot, which are lower than those of the paper by 3.5 Succuess and 4.7 Precision respectively. In our codebase, the Success and Precision are lower 4.8 and 3.7 respectively than those of the paper. Notably, the results below are achieved without hyperparameters search. We observe around 0.5 points fluctuations both in Success and Percision.
+
+|    Backbone     |  Style  | Lr schd | Mem (GB) | Inf time (fps) | Success | Norm Precision | Precision | Config | Download |
+| :-------------: | :-----: | :-----: | :------: | :------------: | :----: | :----: | :------: | :------: | :--------: |
+|    R-50    |  -  |   20e    |  -   | -              | 64.8 | 83.2 | 87.7 | [config](siamese_rpn_r50_1x_otb100.py) | [model](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_otb100/siamese_rpn_r50_1x_otb100_20210920_001757-12636a0a.pth) &#124; [log](https://download.openmmlab.com/mmtracking/sot/siamese_rpn/siamese_rpn_r50_1x_otb100/siamese_rpn_r50_1x_otb100_20210920_001757.log.json) |
diff --git a/configs/sot/siamese_rpn/siamese_rpn_r50_1x_otb100.py b/configs/sot/siamese_rpn/siamese_rpn_r50_1x_otb100.py
@@ -0,0 +1,82 @@
+_base_ = ['./siamese_rpn_r50_1x_lasot.py']
+
+crop_size = 511
+exemplar_size = 127
+search_size = 255
+
+# model settings
+model = dict(
+    test_cfg=dict(rpn=dict(penalty_k=0.4, window_influence=0.5, lr=0.4)))
+
+data_root = 'data/'
+train_pipeline = [
+    dict(type='LoadMultiImagesFromFile', to_float32=True),
+    dict(type='SeqLoadAnnotations', with_bbox=True),
+    dict(
+        type='SeqCropLikeSiamFC',
+        context_amount=0.5,
+        exemplar_size=exemplar_size,
+        crop_size=crop_size),
+    dict(type='SeqGrayAug', prob=0.2),
+    dict(
+        type='SeqShiftScaleAug',
+        target_size=[exemplar_size, search_size],
+        shift=[4, 64],
+        scale=[0.05, 0.18]),
+    dict(type='SeqColorAug', prob=[1.0, 1.0]),
+    dict(type='SeqBlurAug', prob=[0.0, 0.2]),
+    dict(type='VideoCollect', keys=['img', 'gt_bboxes', 'is_positive_pairs']),
+    dict(type='ConcatVideoReferences'),
+    dict(type='SeqDefaultFormatBundle', ref_prefix='search')
+]
+# dataset settings
+data = dict(
+    samples_per_gpu=16,
+    train=[
+        dict(
+            type='RepeatDataset',
+            times=39,
+            dataset=dict(
+                type='SOTTrainDataset',
+                ann_file=data_root +
+                'ILSVRC/annotations/imagenet_vid_train.json',
+                img_prefix=data_root + 'ILSVRC/Data/VID',
+                pipeline=train_pipeline,
+                ref_img_sampler=dict(
+                    frame_range=100,
+                    pos_prob=0.8,
+                    filter_key_img=False,
+                    return_key_img=True),
+            )),
+        dict(
+            type='SOTTrainDataset',
+            ann_file=data_root + 'coco/annotations/instances_train2017.json',
+            img_prefix=data_root + 'coco/train2017',
+            pipeline=train_pipeline,
+            ref_img_sampler=dict(
+                frame_range=0,
+                pos_prob=0.8,
+                filter_key_img=False,
+                return_key_img=True),
+        ),
+        dict(
+            type='SOTTrainDataset',
+            ann_file=data_root +
+            'ILSVRC/annotations/imagenet_det_30plus1cls.json',
+            img_prefix=data_root + 'ILSVRC/Data/DET',
+            pipeline=train_pipeline,
+            ref_img_sampler=dict(
+                frame_range=0,
+                pos_prob=0.8,
+                filter_key_img=False,
+                return_key_img=True),
+        ),
+    ],
+    val=dict(
+        type='OTB100Dataset',
+        ann_file=data_root + 'otb100/annotations/otb100.json',
+        img_prefix=data_root + 'otb100/data'),
+    test=dict(
+        type='OTB100Dataset',
+        ann_file=data_root + 'otb100/annotations/otb100.json',
+        img_prefix=data_root + 'otb100/data'))
diff --git a/configs/sot/siamese_rpn/siamese_rpn_r50_1x_uav123.py b/configs/sot/siamese_rpn/siamese_rpn_r50_1x_uav123.py
@@ -7,6 +7,10 @@
 data_root = 'data/'
 # dataset settings
 data = dict(
+    val=dict(
+        type='UAV123Dataset',
+        ann_file=data_root + 'UAV123/annotations/uav123.json',
+        img_prefix=data_root + 'UAV123/data_seq/UAV123'),
     test=dict(
         type='UAV123Dataset',
         ann_file=data_root + 'UAV123/annotations/uav123.json',

diff --git a/docs/dataset.md b/docs/dataset.md
@@ -10,11 +10,21 @@ This page provides the instructions for dataset preparation on existing benchmar
   - [LaSOT](http://vision.cs.stonybrook.edu/~lasot/)
   - [UAV123](https://cemse.kaust.edu.sa/ivul/uav123/)
   - [TrackingNet](https://tracking-net.org/)
+  - [OTB100](http://www.visual-tracking.net/)
 
 ### 1. Download Datasets
 
 Please download the datasets from the offical websites. It is recommended to symlink the root of the datasets to `$MMTRACKING/data`. If your folder structure is different from the following, you may need to change the corresponding paths in config files.
 
+For OTB100 dataset, you don't need to download the dataset from the official website manually, since we provide a sctipt to download it.
+
+#### OTB100
+
+```shell
+# download OTB100 dataset by web crawling
+python ./tools/convert_datasets/otb100/download_otb100.py -o ./data/otb100/zips -p 8
+```
+
 Notes:
 
 - The `Lists` under `ILSVRC` contains the txt files from [here](https://github.com/msracver/Flow-Guided-Feature-Aggregation/tree/master/data/ILSVRC2015/ImageSets).
@@ -23,7 +33,7 @@ Notes:
 
 - For the training and testing of multi object tracking task, only one of the MOT Challenge dataset (e.g. MOT17) is needed.
 
-- For the training and testing of single object tracking task, the MSCOCO, ILSVRC, LaSOT, UAV123 and TrackingNet datasets are needed.
+- For the training and testing of single object tracking task, the MSCOCO, ILSVRC, LaSOT, UAV123, TrackingNet and OTB100 datasets are needed.
 
 ```
 mmtracking
@@ -77,6 +87,11 @@ mmtracking
 │   │   ├── TEST
 │   │   │   ├── anno
 │   │   │   ├── zips
+│   │
+│   ├── otb100
+│   │   │── zips
+│   │   │   │── Basketball.zip
+│   │   │   │── Biker.zip
 ```
 
 ### 2. Convert Annotations
@@ -103,10 +118,16 @@ python ./tools/convert_datasets/mot/mot2reid.py -i ./data/MOT17/ -o ./data/MOT17
 python ./tools/convert_datasets/uav123/uav2coco.py -i ./data/UAV123/ -o ./data/UAV123/annotations
 
 # TrackingNet
-# unzip files in 'TEST/zips/*.zip'
+# unzip files in 'data/trackingnet/TEST/zips/*.zip'
 bash ./tools/convert_datasets/trackingnet/unzip_trackingnet_test.sh ./data/trackingnet/TEST
 # generate testset annotaions
 python ./tools/convert_datasets/trackingnet/trackingnet2coco.py -i ./data/trackingnet/TEST/ -o ./data/trackingnet/TEST/annotations
+
+# OTB100
+# unzip files in 'data/otb100/zips/*.zip'
+bash ./tools/convert_datasets/otb100/unzip_otb100.sh ./data/otb100
+# generate annotations
+python ./tools/convert_datasets/otb100/otb2coco.py -i ./data/otb100/data -o ./data/otb100/annotations
 ```
 
 The folder structure will be as following after your run these scripts:
@@ -174,6 +195,15 @@ mmtracking
 │   │   │   ├── frames (the unzipped folders)
 │   │   │   │   ├── 0-6LB4FqxoE_0
 │   │   │   │   ├── 07Ysk1C0ZX0_0
+│   │
+│   ├── otb100
+│   │   ├── zips
+│   │   │   ├── Basketball.zip
+│   │   │   ├── Biker.zip
+│   │   ├── annotations
+│   │   ├── data
+│   │   │   ├── Basketball
+│   │   │   │   ├── img
 ```
 
 #### The folder of annotations in ILSVRC
@@ -256,3 +286,11 @@ There are 511 video directories of TrackingNet testset in `data/trackingnet/TEST
 There are only 1 json files in `data/trackingnet/TEST/annotations`:
 
 `trackingnet_test.json`:  Json file containing the annotations information of the testing set in TrackingNet dataset.
+
+#### The folder of data and annotations in OTB100
+
+There are 98 video directories of OTB100 dataset in `data/otb100/data`, and the `img` folder under each video directory contains all images of the video.
+
+There are only 1 json files in `data/otb100/annotations`:
+
+`otb100.json`:  Json file containing the annotations information of the OTB100 dataset.
diff --git a/docs/useful_tools_scripts.md b/docs/useful_tools_scripts.md
@@ -52,7 +52,15 @@ Example on UAV123 dataset:
 ```shell
 ./tools/analysis/sot/dist_sot_siamrpn_param_search.sh [${CONFIG_FILE}] [$GPUS] \
 [--checkpoint ${CHECKPOINT}] [--log ${LOG_FILENAME}] [--eval ${EVAL}] \
-[--penalty-k-range 0.05,0.5,0.05] [--lr-range 0.3,0.45,0.02] [--win-infu-range 0.46,0.55,0.02]
+[--penalty-k-range 0.01,0.22,0.05] [--lr-range 0.4,0.61,0.05] [--win-infu-range 0.01,0.22,0.05]
+```
+
+Example on OTB100 dataset:
+
+```shell
+./tools/analysis/sot/dist_sot_siamrpn_param_search.sh [${CONFIG_FILE}] [$GPUS] \
+[--checkpoint ${CHECKPOINT}] [--log ${LOG_FILENAME}] [--eval ${EVAL}] \
+[--penalty-k-range 0.3,0.45,0.02] [--lr-range 0.35,0.5,0.02] [--win-infu-range 0.46,0.55,0.02]
 ```
 
 ## Log Analysis

diff --git a/docs_zh-CN/dataset.md b/docs_zh-CN/dataset.md
@@ -10,11 +10,21 @@
   - [LaSOT](http://vision.cs.stonybrook.edu/~lasot/)
   - [UAV123](https://cemse.kaust.edu.sa/ivul/uav123/)
   - [TrackingNet](https://tracking-net.org/)
+  - [OTB100](http://www.visual-tracking.net/)
 
 ### 1. 下载数据集
 
 请从官方网站下载数据集。建议将数据集的根目录符号链接到 `$MMTRACKING/data`。如果您的文件夹结构与以下不同，您可能需要更改配置文件中的相应路径。
 
+对于 OTB100 数据集，你不必要手工地从官网下载数据。我们提供了下载脚本。
+
+#### OTB100
+
+```shell
+# 通过网页爬虫下载 OTB100 数据集
+python ./tools/convert_datasets/otb100/download_otb100.py -o ./data/otb100/zips -p 8
+```
+
 注意：
 
 - `ILSVRC` 下的 `Lists` 包含来自在[这里](https://github.com/msracver/Flow-Guided-Feature-Aggregation/tree/master/data/ILSVRC2015/ImageSets)的 txt 文件。
@@ -23,7 +33,7 @@
 
 - 对于多目标跟踪任务的训练和测试，只需要 MOT Challenge 中的任意一个数据集（比如 MOT17）。
 
-- 对于单目标跟踪任务的训练和测试，需要 MSCOCO，ILSVRC, LaSOT, UAV123 和 TrackingNet 数据集。
+- 对于单目标跟踪任务的训练和测试，需要 MSCOCO，ILSVRC, LaSOT, UAV123, TrackingNet 和 OTB100 数据集。
 
 ```
 mmtracking
@@ -77,6 +87,11 @@ mmtracking
 │   │   ├── TEST
 │   │   │   ├── anno
 │   │   │   ├── zips
+│   │
+│   ├── otb100
+│   │   │── zips
+│   │   │   │── Basketball.zip
+│   │   │   │── Biker.zip
 ```
 
 ### 2. 转换标注格式
@@ -104,10 +119,16 @@ python ./tools/convert_datasets/mot/mot2reid.py -i ./data/MOT17/ -o ./data/MOT17
 python ./tools/convert_datasets/uav123/uav2coco.py -i ./data/UAV123/ -o ./data/UAV123/annotations
 
 # TrackingNet
-# 解压目录 'TEST/zips' 下的所有 '*.zip' 文件
+# 解压目录 'data/trackingnet/TEST/zips' 下的所有 '*.zip' 文件
 bash ./tools/convert_datasets/trackingnet/unzip_trackingnet_test.sh ./data/trackingnet/TEST
 # 生成测试集标注
 python ./tools/convert_datasets/trackingnet/trackingnet2coco.py -i ./data/trackingnet/TEST/ -o ./data/trackingnet/TEST/annotations
+
+# OTB100
+# 解压目录 'data/otb100/zips' 下的所有 '*.zip' 文件
+bash ./tools/convert_datasets/otb100/unzip_otb100.sh ./data/otb100
+# 生成标注
+python ./tools/convert_datasets/otb100/otb2coco.py -i ./data/otb100/data -o ./data/otb100/annotations
 ```
 
 完成以上格式转换后，文件目录结构如下：
@@ -175,6 +196,15 @@ mmtracking
 │   │   │   ├── frames (the unzipped folders)
 │   │   │   │   ├── 0-6LB4FqxoE_0
 │   │   │   │   ├── 07Ysk1C0ZX0_0
+│   │
+│   ├── otb100
+│   │   ├── zips
+│   │   │   ├── Basketball.zip
+│   │   │   ├── Biker.zip
+│   │   ├── annotations
+│   │   ├── data
+│   │   │   ├── Basketball
+│   │   │   │   ├── img
 ```
 
 #### ILSVRC的标注文件夹
@@ -258,3 +288,11 @@ MOT17-02-FRCNN_000009/000081.jpg 3
 在 `data/trackingnet/TEST/annotations` 中只有一个 json 文件：
 
 `trackingnet_test.json`： 包含 TrackingNet 测试集标注信息的 json 文件。
+
+#### TrackingNet的标注和视频帧文件夹
+
+在 `data/otb100/data` 文件夹下有 OTB100 数据集的 98 个视频目录， 每个视频目录下的 `img` 文件夹包含该视频所有图片。
+
+在 `data/otb100/data/annotations` 中只有一个 json 文件：
+
+`otb100.json`： 包含 OTB100 数据集标注信息的 json 文件
diff --git a/docs_zh-CN/useful_tools_scripts.md b/docs_zh-CN/useful_tools_scripts.md
@@ -52,7 +52,15 @@
 ```shell
 ./tools/analysis/sot/dist_sot_siamrpn_param_search.sh [${CONFIG_FILE}] [$GPUS] \
 [--checkpoint ${CHECKPOINT}] [--log ${LOG_FILENAME}] [--eval ${EVAL}] \
-[--penalty-k-range 0.05,0.5,0.05] [--lr-range 0.3,0.45,0.02] [--win-infu-range 0.46,0.55,0.02]
+[--penalty-k-range 0.01,0.22,0.05] [--lr-range 0.4,0.61,0.05] [--win-infu-range 0.01,0.22,0.05]
+```
+
+在 OTB100 上的超参搜索范例：
+
+```shell
+./tools/analysis/sot/dist_sot_siamrpn_param_search.sh [${CONFIG_FILE}] [$GPUS] \
+[--checkpoint ${CHECKPOINT}] [--log ${LOG_FILENAME}] [--eval ${EVAL}] \
+[--penalty-k-range 0.3,0.45,0.02] [--lr-range 0.35,0.5,0.02] [--win-infu-range 0.46,0.55,0.02]
 ```
 
 ## 日志分析

diff --git a/mmtrack/datasets/__init__.py b/mmtrack/datasets/__init__.py
@@ -6,6 +6,7 @@
 from .imagenet_vid_dataset import ImagenetVIDDataset
 from .lasot_dataset import LaSOTDataset
 from .mot_challenge_dataset import MOTChallengeDataset
+from .otb_dataset import OTB100Dataset
 from .parsers import CocoVID
 from .pipelines import PIPELINES
 from .reid_dataset import ReIDDataset
@@ -18,5 +19,5 @@
     'DATASETS', 'PIPELINES', 'build_dataloader', 'build_dataset', 'CocoVID',
     'CocoVideoDataset', 'ImagenetVIDDataset', 'MOTChallengeDataset',
     'ReIDDataset', 'SOTTrainDataset', 'SOTTestDataset', 'LaSOTDataset',
-    'UAV123Dataset', 'TrackingNetTestDataset'
+    'UAV123Dataset', 'TrackingNetTestDataset', 'OTB100Dataset'
 ]
diff --git a/mmtrack/datasets/otb_dataset.py b/mmtrack/datasets/otb_dataset.py
@@ -0,0 +1,11 @@
+from mmtrack.datasets import DATASETS
+from .sot_test_dataset import SOTTestDataset
+
+
+@DATASETS.register_module()
+class OTB100Dataset(SOTTestDataset):
+    """OTB100 dataset for the testing of single object tracking.
+
+    The dataset doesn't support training mode.
+    """
+    pass
diff --git a/mmtrack/datasets/pipelines/__init__.py b/mmtrack/datasets/pipelines/__init__.py
@@ -7,15 +7,15 @@
                       SeqLoadAnnotations)
 from .processing import MatchInstances
 from .transforms import (SeqBlurAug, SeqColorAug, SeqCropLikeSiamFC,
-                         SeqNormalize, SeqPad, SeqPhotoMetricDistortion,
-                         SeqRandomCrop, SeqRandomFlip, SeqResize,
-                         SeqShiftScaleAug)
+                         SeqGrayAug, SeqNormalize, SeqPad,
+                         SeqPhotoMetricDistortion, SeqRandomCrop,
+                         SeqRandomFlip, SeqResize, SeqShiftScaleAug)
 
 __all__ = [
     'PIPELINES', 'LoadMultiImagesFromFile', 'SeqLoadAnnotations', 'SeqResize',
     'SeqNormalize', 'SeqRandomFlip', 'SeqPad', 'SeqDefaultFormatBundle',
     'VideoCollect', 'ConcatVideoReferences', 'LoadDetections',
     'MatchInstances', 'SeqRandomCrop', 'SeqPhotoMetricDistortion',
     'SeqCropLikeSiamFC', 'SeqShiftScaleAug', 'SeqBlurAug', 'SeqColorAug',
-    'ToList', 'ReIDFormatBundle'
+    'ToList', 'ReIDFormatBundle', 'SeqGrayAug'
 ]