Add Biwi Kinect Head Pose dataset. (#3903)

* Initial draft for Biwi Kinect Head Pose dataset.

* fix typo.

* update dataset features and complete generate_examples.

* add dummy data.

* add dataset_infos.json

* update Dataset Structure section.

* update Readme.

* add data processing section, sample code to read a compressed binary depth image.

* add size category info.

* Update datasets/biwi_kinect_head_pose/biwi_kinect_head_pose.py

Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

* Update datasets/biwi_kinect_head_pose/biwi_kinect_head_pose.py

Use os.path.join for consistency.

Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

* Update datasets/biwi_kinect_head_pose/biwi_kinect_head_pose.py

Use os.path.join

Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

* Update datasets/biwi_kinect_head_pose/biwi_kinect_head_pose.py

Add encoding when we open files.

Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

* Update datasets/biwi_kinect_head_pose/biwi_kinect_head_pose.py

Add encoding when we open files.

Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

* add missing quote.

* collapse C++ code.

* add full name to pretty_name

* new dataset_infos.json file.

* replace tabs with spaces.

* Update datasets/biwi_kinect_head_pose/README.md

Add contributions info.

Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

Commit from https://github.com/huggingface/datasets/commit/55e6b308c6b6ee039b33752a1b464fb9f934f1cd

README.md

@@ -0,0 +1,308 @@
+---
+annotations_creators:
+- expert-generated
+language_creators:
+- expert-generated
+languages:
+- en
+licenses:
+- other
+multilinguality:
+- monolingual
+pretty_name: Biwi Kinect Head Pose Database
+size_categories:
+- 10K<n<100K
+source_datasets:
+- original
+task_categories:
+- other
+task_ids:
+- other-other-head-pose-estimation
+paperswithcode_id: biwi
+---
+
+# Dataset Card for Biwi Kinect Head Pose Database
+
+## Table of Contents
+- [Dataset Description](#dataset-description)
+  - [Dataset Summary](#dataset-summary)
+  - [Supported Tasks](#supported-tasks-and-leaderboards)
+  - [Languages](#languages)
+- [Dataset Structure](#dataset-structure)
+  - [Data Instances](#data-instances)
+  - [Data Fields](#data-instances)
+  - [Data Splits](#data-instances)
+- [Dataset Creation](#dataset-creation)
+  - [Curation Rationale](#curation-rationale)
+  - [Source Data](#source-data)
+  - [Annotations](#annotations)
+  - [Personal and Sensitive Information](#personal-and-sensitive-information)
+- [Considerations for Using the Data](#considerations-for-using-the-data)
+  - [Social Impact of Dataset](#social-impact-of-dataset)
+  - [Discussion of Biases](#discussion-of-biases)
+  - [Other Known Limitations](#other-known-limitations)
+- [Additional Information](#additional-information)
+  - [Dataset Curators](#dataset-curators)
+  - [Licensing Information](#licensing-information)
+  - [Citation Information](#citation-information)
+
+## Dataset Description
+
+- **Homepage:** [Biwi Kinect Head Pose homepage](https://icu.ee.ethz.ch/research/datsets.html)
+- **Repository:** [Needs More Information]
+- **Paper:** [Biwi Kinect Head Pose paper](https://link.springer.com/article/10.1007/s11263-012-0549-0)
+- **Leaderboard:** [Needs More Information]
+- **Point of Contact:** [Gabriele Fanelli](mailto:gabriele.fanelli@gmail.com)
+
+### Dataset Summary
+
+The Biwi Kinect Head Pose Database is acquired with the Microsoft Kinect sensor, a structured IR light device.It contains 15K images of 20 people with 6 females and 14 males where 4 people were recorded twice.
+
+For each frame, there is :
+- a depth image,
+- a corresponding rgb image (both 640x480 pixels),
+- annotation
+
+The head pose range covers about +-75 degrees yaw and +-60 degrees pitch. The ground truth is the 3D location of the head and its rotation.
+
+### Data Processing
+
+Example code for reading a compressed binary depth image file provided by the authors.
+
+<details>
+  <summary> View C++ Code </summary>
+
+```cpp
+/*
+ * Gabriele Fanelli
+ *
+ * fanelli@vision.ee.ethz.ch
+ *
+ * BIWI, ETHZ, 2011
+ *
+ * Part of the Biwi Kinect Head Pose Database
+ *
+ * Example code for reading a compressed binary depth image file.
+ *
+ * THE SOFTWARE IS PROVIDED “AS IS” AND THE PROVIDER GIVES NO EXPRESS OR IMPLIED WARRANTIES OF ANY KIND,
+ * INCLUDING WITHOUT LIMITATION THE WARRANTIES OF FITNESS FOR ANY PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL THE PROVIDER BE HELD RESPONSIBLE FOR LOSS OR DAMAGE CAUSED BY THE USE OF THE SOFTWARE.
+ *
+ *
+ */
+
+#include <iostream>
+#include <fstream>
+#include <cstdlib>
+
+int16_t* loadDepthImageCompressed( const char* fname ){
+
+    //now read the depth image
+    FILE* pFile = fopen(fname, "rb");
+    if(!pFile){
+        std::cerr << "could not open file " << fname << std::endl;
+        return NULL;
+    }
+
+    int im_width = 0;
+    int im_height = 0;
+    bool success = true;
+
+    success &= ( fread(&im_width,sizeof(int),1,pFile) == 1 ); // read width of depthmap
+    success &= ( fread(&im_height,sizeof(int),1,pFile) == 1 ); // read height of depthmap
+
+    int16_t* depth_img = new int16_t[im_width*im_height];
+    
+    int numempty;
+    int numfull;
+    int p = 0;
+
+    while(p < im_width*im_height ){
+
+        success &= ( fread( &numempty,sizeof(int),1,pFile) == 1 );
+
+        for(int i = 0; i < numempty; i++)
+            depth_img[ p + i ] = 0;
+
+        success &= ( fread( &numfull,sizeof(int), 1, pFile) == 1 );
+        success &= ( fread( &depth_img[ p + numempty ], sizeof(int16_t), numfull, pFile) == (unsigned int) numfull );
+        p += numempty+numfull;
+
+    }
+
+    fclose(pFile);
+
+    if(success)
+        return depth_img;
+    else{
+        delete [] depth_img;
+        return NULL;
+    }
+}
+
+float* read_gt(const char* fname){
+
+    //try to read in the ground truth from a binary file
+    FILE* pFile = fopen(fname, "rb");
+    if(!pFile){
+        std::cerr << "could not open file " << fname << std::endl;
+        return NULL;
+    }
+    
+    float* data = new float[6];
+    
+    bool success = true;
+    success &= ( fread( &data[0], sizeof(float), 6, pFile) == 6 );
+    fclose(pFile);
+    
+    if(success)
+        return data;
+    else{
+        delete [] data;
+        return NULL;
+    }
+
+}
+```
+
+</details>
+
+
+### Supported Tasks and Leaderboards
+
+Biwi Kinect Head Pose Database supports the following tasks :
+- Head pose estimation
+- Pose estimation
+- Face verification
+
+### Languages
+
+[Needs More Information]
+
+## Dataset Structure
+
+### Data Instances
+
+A sample from the Biwi Kinect Head Pose dataset is provided below:
+
+```
+{
+    'sequence_number': '12', 
+    'subject_id': 'M06', 
+    'rgb': [<PIL.PngImagePlugin.PngImageFile image mode=RGB size=640x480 at 0x7F53A6446C10>,.....],
+    'rgb_cal': 
+        {
+            'intrisic_mat': [[517.679, 0.0, 320.0], [0.0, 517.679, 240.5], [0.0, 0.0, 1.0]],
+            'extrinsic_mat': 
+            {
+                'rotation': [[0.999947, 0.00432361, 0.00929419], [-0.00446314, 0.999877, 0.0150443], [-0.009228, -0.015085, 0.999844]], 
+                'translation': [-24.0198, 5.8896, -13.2308]
+            }
+        }
+    'depth': ['../hpdb/12/frame_00003_depth.bin', .....],
+    'depth_cal': 
+        {
+            'intrisic_mat': [[575.816, 0.0, 320.0], [0.0, 575.816, 240.0], [0.0, 0.0, 1.0]],
+            'extrinsic_mat': 
+            {
+                'rotation': [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], 
+                'translation': [0.0, 0.0, 0.0]
+            }
+        }
+    'head_pose_gt': 
+        {
+            'center': [[43.4019, -30.7038, 906.864], [43.0202, -30.8683, 906.94], [43.0255, -30.5611, 906.659], .....],
+            'rotation': [[[0.980639, 0.109899, 0.162077], [-0.11023, 0.993882, -0.00697376], [-0.161851, -0.011027, 0.986754]], ......]
+        }
+}
+```
+
+### Data Fields
+
+- `sequence_number` : This refers to the sequence number in the dataset. There are a total of 24 sequences.
+- `subject_id` : This refers to the subjects in the dataset. There are a total of 20 people with 6 females and 14 males where 4 people were recorded twice.
+- `rgb` : List of png frames containing the poses.
+- `rgb_cal`: Contains calibration information for the color camera which includes intrinsic matrix, 
+global rotation and translation.
+- `depth` : List of depth frames for the poses.
+- `depth_cal`: Contains calibration information for the depth camera which includes intrinsic matrix, global rotation and translation.
+- `head_pose_gt` : Contains ground truth information, i.e., the location of the center of the head in 3D and the head rotation, encoded as a 3x3 rotation matrix.
+
+
+### Data Splits
+
+All the data is contained in the training set.
+
+## Dataset Creation
+
+### Curation Rationale
+
+[More Information Needed]
+
+### Source Data
+
+#### Initial Data Collection and Normalization
+
+The Biwi Kinect Head Pose Database is acquired with the Microsoft Kinect sensor, a structured IR light device.
+#### Who are the source language producers?
+
+[More Information Needed]
+
+### Annotations
+
+#### Annotation process
+
+From Dataset's README : 
+> The database contains 24 sequences acquired with a Kinect sensor. 20 people (some were recorded twice - 6 women and 14 men) were recorded while turning their heads, sitting in front of the sensor, at roughly one meter of distance.
+
+#### Who are the annotators?
+
+[More Information Needed]
+
+### Personal and Sensitive Information
+
+[More Information Needed]
+
+## Considerations for Using the Data
+
+### Social Impact of Dataset
+
+[More Information Needed]
+
+### Discussion of Biases
+
+[More Information Needed]
+
+### Other Known Limitations
+
+[More Information Needed]
+
+## Additional Information
+
+### Dataset Curators
+
+[Needs More Information]
+
+### Licensing Information
+
+From Dataset's README : 
+> This database is made available for non-commercial use such as university research and education.
+
+### Citation Information
+
+```bibtex
+@article{fanelli_IJCV,
+  author = {Fanelli, Gabriele and Dantone, Matthias and Gall, Juergen and Fossati, Andrea and Van Gool, Luc},
+  title = {Random Forests for Real Time 3D Face Analysis},
+  journal = {Int. J. Comput. Vision},
+  year = {2013},
+  month = {February},
+  volume = {101},
+  number = {3},
+  pages = {437--458}
+}
+```
+
+### Contributions
+
+Thanks to [@dnaveenr](https://github.com/dnaveenr) for adding this dataset.

biwi_kinect_head_pose.py

@@ -0,0 +1,215 @@
+# Copyright 2022 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Biwi Kinect Head Pose Database."""
+
+
+import glob
+import os
+
+import datasets
+
+
+_CITATION = """\
+@article{fanelli_IJCV,
+  author = {Fanelli, Gabriele and Dantone, Matthias and Gall, Juergen and Fossati, Andrea and Van Gool, Luc},
+  title = {Random Forests for Real Time 3D Face Analysis},
+  journal = {Int. J. Comput. Vision},
+  year = {2013},
+  month = {February},
+  volume = {101},
+  number = {3},
+  pages = {437--458}
+}
+"""
+
+
+_DESCRIPTION = """\
+The Biwi Kinect Head Pose Database is acquired with the Microsoft Kinect sensor, a structured IR light device.It contains 15K images of 20 people with 6 females and 14 males where 4 people were recorded twice.
+"""
+
+
+_HOMEPAGE = "https://icu.ee.ethz.ch/research/datsets.html"
+
+
+_LICENSE = "This database is made available for non-commercial use such as university research and education."
+
+
+_URLS = {
+    "kinect_head_pose_db": "https://data.vision.ee.ethz.ch/cvl/gfanelli/kinect_head_pose_db.tgz",
+}
+
+_sequence_to_subject_map = {
+    "01": "F01",
+    "02": "F02",
+    "03": "F03",
+    "04": "F04",
+    "05": "F05",
+    "06": "F06",
+    "07": "M01",
+    "08": "M02",
+    "09": "M03",
+    "10": "M04",
+    "11": "M05",
+    "12": "M06",
+    "13": "M07",
+    "14": "M08",
+    "15": "F03",
+    "16": "M09",
+    "17": "M10",
+    "18": "F05",
+    "19": "M11",
+    "20": "M12",
+    "21": "F02",
+    "22": "M01",
+    "23": "M13",
+    "24": "M14",
+}
+
+
+class BiwiKinectHeadPose(datasets.GeneratorBasedBuilder):
+
+    VERSION = datasets.Version("1.0.0")
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "sequence_number": datasets.Value("string"),
+                    "subject_id": datasets.Value("string"),
+                    "rgb": datasets.Sequence(datasets.Image()),
+                    "rgb_cal": {
+                        "intrisic_mat": datasets.Array2D(shape=(3, 3), dtype="float64"),
+                        "extrinsic_mat": {
+                            "rotation": datasets.Array2D(shape=(3, 3), dtype="float64"),
+                            "translation": datasets.Sequence(datasets.Value("float64"), length=3),
+                        },
+                    },
+                    "depth": datasets.Sequence(datasets.Value("string")),
+                    "depth_cal": {
+                        "intrisic_mat": datasets.Array2D(shape=(3, 3), dtype="float64"),
+                        "extrinsic_mat": {
+                            "rotation": datasets.Array2D(shape=(3, 3), dtype="float64"),
+                            "translation": datasets.Sequence(datasets.Value("float64"), length=3),
+                        },
+                    },
+                    "head_pose_gt": datasets.Sequence(
+                        {
+                            "center": datasets.Sequence(datasets.Value("float64"), length=3),
+                            "rotation": datasets.Array2D(shape=(3, 3), dtype="float64"),
+                        }
+                    ),
+                    "head_template": datasets.Value("string"),
+                }
+            ),
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+
+        data_dir = dl_manager.download_and_extract(_URLS)
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "dataset_path": os.path.join(data_dir["kinect_head_pose_db"], "hpdb"),
+                },
+            ),
+        ]
+
+    @staticmethod
+    def _get_calibration_information(cal_file_path):
+        with open(cal_file_path, "r", encoding="utf-8") as f:
+            cal_info = f.read().splitlines()
+
+        intrisic_mat = []
+        extrinsic_mat = []
+
+        for data in cal_info[:3]:
+            row = list(map(float, data.strip().split(" ")))
+            intrisic_mat.append(row)
+
+        for data in cal_info[6:9]:
+            row = list(map(float, data.strip().split(" ")))
+            extrinsic_mat.append(row)
+
+        translation = list(map(float, cal_info[10].strip().split(" ")))
+
+        return {
+            "intrisic_mat": intrisic_mat,
+            "extrinsic_mat": {
+                "rotation": extrinsic_mat,
+                "translation": translation,
+            },
+        }
+
+    @staticmethod
+    def _parse_head_pose_info(head_pose_file):
+        with open(head_pose_file, "r", encoding="utf-8") as f:
+            head_pose_info = f.read().splitlines()
+
+        rotation = []
+        for data in head_pose_info[:3]:
+            row = list(map(float, data.strip().split(" ")))
+            rotation.append(row)
+
+        center = list(map(float, head_pose_info[4].strip().split(" ")))
+
+        return {
+            "center": center,
+            "rotation": rotation,
+        }
+
+    @staticmethod
+    def _get_head_pose_information(path):
+        head_pose_files = sorted(glob.glob(os.path.join(path, "*_pose.txt")))
+
+        head_poses_info = []
+
+        for head_pose_file in head_pose_files:
+            head_pose = BiwiKinectHeadPose._parse_head_pose_info(head_pose_file)
+            head_poses_info.append(head_pose)
+
+        return head_poses_info
+
+    def _generate_examples(self, dataset_path):
+
+        idx = 0
+        folders = os.listdir(dataset_path)
+        for item in folders:
+            sequence_number = item
+            sequence_base_path = os.path.join(dataset_path, sequence_number)
+            if os.path.isdir(sequence_base_path):
+                rgb_files = sorted(glob.glob(os.path.join(sequence_base_path, "*.png")))
+                depth_files = sorted(glob.glob(os.path.join(sequence_base_path, "*.bin")))
+                head_template_path = os.path.join(dataset_path, sequence_number + ".obj")
+                rgb_cal = self._get_calibration_information(os.path.join(sequence_base_path, "rgb.cal"))
+                depth_cal = self._get_calibration_information(os.path.join(sequence_base_path, "depth.cal"))
+                head_pose_gt = self._get_head_pose_information(sequence_base_path)
+
+                yield idx, {
+                    "sequence_number": sequence_number,
+                    "subject_id": _sequence_to_subject_map[sequence_number],
+                    "rgb": rgb_files,
+                    "rgb_cal": rgb_cal,
+                    "depth": depth_files,
+                    "depth_cal": depth_cal,
+                    "head_pose_gt": head_pose_gt,
+                    "head_template": head_template_path,
+                }
+
+                idx += 1

dataset_infos.json

@@ -0,0 +1 @@
+{"default": {"description": "The Biwi Kinect Head Pose Database is acquired with the Microsoft Kinect sensor, a structured IR light device.It contains 15K images of 20 people with 6 females and 14 males where 4 people were recorded twice.\n", "citation": "@article{fanelli_IJCV,\n  author = {Fanelli, Gabriele and Dantone, Matthias and Gall, Juergen and Fossati, Andrea and Van Gool, Luc},\n  title = {Random Forests for Real Time 3D Face Analysis},\n  journal = {Int. J. Comput. Vision},\n  year = {2013},\n  month = {February},\n  volume = {101},\n  number = {3},\n  pages = {437--458}\n}\n", "homepage": "https://icu.ee.ethz.ch/research/datsets.html", "license": "This database is made available for non-commercial use such as university research and education.", "features": {"sequence_number": {"dtype": "string", "id": null, "_type": "Value"}, "subject_id": {"dtype": "string", "id": null, "_type": "Value"}, "rgb": {"feature": {"decode": true, "id": null, "_type": "Image"}, "length": -1, "id": null, "_type": "Sequence"}, "rgb_cal": {"intrisic_mat": {"shape": [3, 3], "dtype": "float64", "id": null, "_type": "Array2D"}, "extrinsic_mat": {"rotation": {"shape": [3, 3], "dtype": "float64", "id": null, "_type": "Array2D"}, "translation": {"feature": {"dtype": "float64", "id": null, "_type": "Value"}, "length": 3, "id": null, "_type": "Sequence"}}}, "depth": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "depth_cal": {"intrisic_mat": {"shape": [3, 3], "dtype": "float64", "id": null, "_type": "Array2D"}, "extrinsic_mat": {"rotation": {"shape": [3, 3], "dtype": "float64", "id": null, "_type": "Array2D"}, "translation": {"feature": {"dtype": "float64", "id": null, "_type": "Value"}, "length": 3, "id": null, "_type": "Sequence"}}}, "head_pose_gt": {"feature": {"center": {"feature": {"dtype": "float64", "id": null, "_type": "Value"}, "length": 3, "id": null, "_type": "Sequence"}, "rotation": {"shape": [3, 3], "dtype": "float64", "id": null, "_type": "Array2D"}}, "length": -1, "id": null, "_type": "Sequence"}, "head_template": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "biwi_kinect_head_pose", "config_name": "default", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 6914063, "num_examples": 24, "dataset_name": "biwi_kinect_head_pose"}}, "download_checksums": {"https://data.vision.ee.ethz.ch/cvl/gfanelli/kinect_head_pose_db.tgz": {"num_bytes": 6014398431, "checksum": "d8fc0fee11b6b865b18b292de7c21dd2181492bd770c4fe13821e8dc630f5549"}}, "download_size": 6014398431, "post_processing_size": null, "dataset_size": 6914063, "size_in_bytes": 6021312494}}

dummy/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58468f9f93a1fd98c8f2e1f702d717bfbbbf968f8d8f78744ec1330abd8a8459
+size 829287