Update README.md

README.md

@@ -1,10 +1,224 @@
 ---
-title: README
-emoji: 🏆
-colorFrom: purple
-colorTo: gray
-sdk: static
-pinned: false
+task_categories:
+- image-segmentation
+- object-detection
+- robotics
+- zero-shot-object-detection
+size_categories:
+- n>1T
+configs:
+- config_name: MegaPose-ShapeNetCore
+  data_files: MegaPose-ShapeNetCore/*.tar
+- config_name: MegaPose-GSO
+  data_files: MegaPose-GSO/*.tar
 ---
+# BOP: Benchmark for 6D Object Pose Estimation
+The goal of BOP is to capture the state of the art in estimating the 6D pose, i.e. 3D translation and 3D rotation, of rigid objects from RGB/RGB-D images. An accurate, fast, robust, scalable and easy-to-train method that solves this task will have a big impact in application fields such as robotics or augmented reality.
 
-Edit this `README.md` markdown file to author your organization card.
+![image/jpeg](https://cdn-uploads.huggingface.co/production/uploads/637fb712084fca81acde6e40/8WSyi9CNNsfDHC-lwaRpG.jpeg)
+
+Homepage: https://bop.felk.cvut.cz/home/
+
+Toolkit: https://github.com/thodan/bop_toolkit
+
+## Downloading datasets
+
+#### Option 1: Using `huggingface_hub`:
+
+<details><summary>Click to expand</summary>
+
+a. Install the library:
+```
+pip install --upgrade huggingface_hub
+```
+b. Download the dataset:
+```
+from huggingface_hub import snapshot_download
+
+dataset_name = "hope"
+local_dir = "./datasets"
+
+snapshot_download(repo_id="bop-benchmark/datasets", 
+                  allow_patterns=f"{dataset_name}/*zip",
+                  repo_type="dataset", 
+                  local_dir=local_dir)
+```
+If you want to download the entire BOP datasets (~3TB), please remove the `allow_patterns` argument. More options are available in the [official documentation](https://huggingface.co/docs/huggingface_hub/main/en/guides/download).
+
+</details>
+
+
+#### Option 2: Using `huggingface_hub[cli]`:
+
+<details><summary>Click to expand</summary>
+
+a. Install the library:
+```
+pip install -U "huggingface_hub[cli]"
+```
+b. Download the dataset:
+```
+export LOCAL_DIR=./datasets
+export DATASET_NAME=hope
+
+huggingface-cli download bop-benchmark/datasets --include "$DATASET_NAME/*.zip" --local-dir $LOCAL_DIR --repo-type=dataset  
+```
+Please remove this argument `--include "$DATASET_NAME/*.zip"` to download entire BOP datasets (~3TB). More options are available in the [official documentation](https://huggingface.co/docs/huggingface_hub/main/en/guides/download).
+</details>
+
+#### Option 3: Using `wget`:
+
+<details><summary>Click to expand</summary>
+
+Similar `wget` command as in [BOP website](https://bop.felk.cvut.cz/datasets/) can be used to download the dataset from huggingface hub:
+```
+export SRC=https://huggingface.co/datasets/bop-benchmark/datasets/resolve/main
+
+wget $SRC/lm/lm_base.zip         # Base archive 
+wget $SRC/lm/lm_models.zip       # 3D object models
+wget $SRC/lm/lm_test_all.zip     # All test images ("_bop19" for a subset)
+wget $SRC/lm/lm_train_pbr.zip    # PBR training images 
+```
+</details>
+
+Datasets are stored in `.zip` format. You can extract them using the following command:
+```
+bash scripts/extract_bop.sh
+```
+
+If you are running on a machine with high bandwidth, you can increase your download speed by adding the following environment variable:
+```
+pip install huggingface_hub[hf_transfer]
+export HF_HUB_ENABLE_HF_TRANSFER=1
+```
+
+## Uploading datasets
+
+You create a new dataset and want to share it with BOP community. Here is a step-by-step guide to upload the dataset and create a pull request to [our huggingface hub](https://huggingface.co/datasets/bop-benchmark/datasets/). Feel free to reach out to vanngn.nguyen@gmail.com if you have any questions.
+
+Similar to the download process, you can upload the dataset using the `huggingface_hub` library or `huggingface_hub[cli]`. We recommend using `huggingface_hub[cli]` for its simplicity.
+
+#### Option 1: Using `huggingface_hub[cli]`:
+
+<details><summary>Click to expand</summary>
+
+a. Install the library:
+```
+pip install -U "huggingface_hub[cli]"
+```
+
+b. Log-in and create a token
+```
+huggingface-cli login
+```
+Then go to [this link](https://huggingface.co/settings/tokens) and generate a token. IMPORTANT: the token should have write access as shown below:
+
+<img src="./media/token_hf.png" alt="image" width="300">
+
+
+Make sure you are in the bop-benchmark group by running:
+```
+huggingface-cli whoami
+```
+
+c. Upload dataset:
+
+The command is applied for both folders and specific files:
+```
+# Usage:  huggingface-cli upload bop-benchmark/datasets [local_path] [path_in_repo] --repo-type=dataset --create-pr
+```
+For example, to upload hope dataset:
+```
+export LOCAL_FOLDER=./datasets/hope
+export HF_FOLDER=/hope
+
+huggingface-cli upload bop-benchmark/datasets $LOCAL_FOLDER $HF_FOLDER --repo-type=dataset --create-pr
+```
+
+</details>
+
+#### Option 2: Using `huggingface_hub`:
+
+<details><summary>Click to expand</summary>
+
+a. Install the library:
+```
+pip install --upgrade huggingface_hub
+```
+b. Creating a pull-request:
+
+We recommend organizing the dataset in a folder and then uploading it to the huggingface hub. For example, to upload `lmo`:
+```
+from huggingface_hub import HfApi, CommitOperationAdd
+
+dataset_name = "lmo"
+local_dir = "./datasets/lmo"
+
+operations = []
+for file in local_dir.glob("*"):
+    add_commit = CommitOperationAdd(
+        path_in_repo=f"/{dataset_name}",
+        path_or_fileobj=local_dir,
+    )
+    operations.append(add_commit)
+
+
+api = HfApi()
+MY_TOKEN = # get from https://huggingface.co/settings/tokens
+api.create_commit(repo_id="bop-benchmark/datasets", 
+                  repo_type="dataset",
+                  commit_message=f"adding {dataset_name} dataset", 
+                  token=MY_TOKEN,
+                  operations=operations, 
+                  create_pr=True)
+
+```
+If your dataset is large (> 500 GB), you can upload it in chunks by adding the `multi_commits=True, multi_commits_verbose=True,` argument. More options are available in the [official documentation](https://huggingface.co/docs/huggingface_hub/v0.22.2/en/package_reference/hf_api#huggingface_hub.HfApi.create_pull_request).
+
+</details>
+
+## FAQ
+
+#### 1. How to upload a large file > 50 GB?
+Note that HuggingFace limits the size of each file to 50 GB. If your dataset is larger, you can split it into smaller files:
+```
+zip -s 50g input.zip --out output.zip
+```
+This command will split the `input.zip` into multiple files of 50GB size `output.zip`, `output.z01`, `output.z01`, ... You can then extract them using one of the following commands:
+```
+# option 1: combine 
+zip -s0 output.zip --out input.zip
+
+# option 2: using 7z to unzip directly
+7z x output.zip
+```
+#### 2. How to increase download speed?
+If you are running on a machine with high bandwidth, you can increase your download speed by adding the following environment variable:
+```
+pip install huggingface_hub[hf_transfer]
+export HF_HUB_ENABLE_HF_TRANSFER=1
+```
+
+## Publications
+- [**BOP Challenge 2023 on Detection, Segmentation and Pose Estimation of Seen and Unseen Rigid Objects**](https://arxiv.org/pdf/2403.09799.pdf)
+  - T. Hodaň, M. Sundermeyer, Y. Labbé, V. N. Nguyen, G. Wang, E. Brachmann, B. Drost, V. Lepetit, C. Rother, J. Matas
+  - IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW, [CV4MR workshop](https://cv4mr.github.io/)) 2024, Seattle
+  - [PDF](https://arxiv.org/pdf/2403.09799.pdf), [SLIDES](https://cmp.felk.cvut.cz/sixd/workshop_2023/slides/bop_challenge_2023_results.pdf), [VIDEO](https://www.youtube.com/watch?v=PcDszFANcDQ), [BIB](https://cmp.felk.cvut.cz/~hodanto2/data/hodan2023bop.bib)
+
+- [**BOP Challenge 2022 on Detection, Segmentation and Pose Estimation of Specific Rigid Objects**](https://arxiv.org/pdf/2302.13075.pdf)
+  - M. Sundermeyer, T. Hodaň, Y. Labbé, G. Wang, E. Brachmann, B. Drost, C. Rother, J. Matas
+  - IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW, [CV4MR workshop](https://cv4mr.github.io/)) 2023, Vancouver
+  - [PDF](https://arxiv.org/pdf/2302.13075.pdf), [SLIDES](https://cmp.felk.cvut.cz/sixd/workshop_2022/slides/bop_challenge_2022_results.pdf), [VIDEO 1](https://vimeo.com/showcase/9946695/video/768457697), [VIDEO 2](https://vimeo.com/showcase/9946695/video/768458355), [BIB](https://cmp.felk.cvut.cz/~hodanto2/data/sundermeyer2022bop.bib)
+
+- [**BOP Challenge 2020 on 6D Object Localization**](https://arxiv.org/pdf/2009.07378.pdf)
+  - T. Hodaň, M. Sundermeyer, B. Drost, Y. Labbé, E. Brachmann, F. Michel, C. Rother, J. Matas
+  - European Conference on Computer Vision Workshops (ECCVW) 2020, Glasgow
+  - [PDF](https://arxiv.org/pdf/2009.07378.pdf), [SLIDES](https://bop.felk.cvut.cz/media/bop_challenge_2020_results.pdf), [BIB](http://cmp.felk.cvut.cz/~hodanto2/data/hodan2020bop.bib)
+
+- [**BOP: Benchmark for 6D Object Pose Estimation**](http://cmp.felk.cvut.cz/~hodanto2/data/hodan2018bop.pdf)
+  - T. Hodaň, F. Michel, E. Brachmann, W. Kehl, A. G. Buch, D. Kraft, B. Drost, J. Vidal, S. Ihrke, X. Zabulis, C. Sahin, F. Manhardt, F. Tombari, T.-K. Kim, J. Matas, C. Rother
+  - European Conference on Computer Vision (ECCV) 2018, Munich
+  - [PDF](http://cmp.felk.cvut.cz/~hodanto2/data/hodan2018bop.pdf), [SLIDES](http://cmp.felk.cvut.cz/~hodanto2/data/hodan2018bop_slides_eccv.pdf), [POSTER](http://cmp.felk.cvut.cz/~hodanto2/data/hodan2018bop_poster.pdf), [BIB](http://cmp.felk.cvut.cz/~hodanto2/data/hodan2018bop.bib)
+  
+
+The online evaluation system has been developed by [T. Hodaň](http://www.hodan.xyz) and [A. Melenovský](https://www.linkedin.com/in/anton%C3%ADn-melenovsk%C3%BD-09907b151/).