Init ego4d

.gitattributes

@@ -0,0 +1,37 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text

dataset_infos.json

@@ -0,0 +1 @@
+{"default": {"description": "EGO4D is the world's largest egocentric (first person) video ML dataset and benchmark suite, with 3,600 hrs (and counting) of densely narrated video and a wide range of annotations across five new benchmark tasks. It covers hundreds of scenarios (household, outdoor, workplace, leisure, etc.) of daily life activity captured in-the-wild by 926 unique camera wearers from 74 worldwide locations and 9 different countries. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. The approach to data collection was designed to uphold rigorous privacy and ethics standards with consenting participants and robust de-identification procedures where relevant.\n", "citation": "\n@article{Ego4D2021,\n  author={Grauman, Kristen and Westbury, Andrew and Byrne, Eugene and Chavis, Zachary and Furnari, Antonino and Girdhar, Rohit and Hamburger, Jackson and Jiang, Hao and Liu, Miao and Liu, Xingyu and Martin, Miguel and Nagarajan, Tushar and Radosavovic, Ilija and Ramakrishnan, Santhosh Kumar and Ryan, Fiona and Sharma, Jayant and Wray, Michael and Xu, Mengmeng and Xu, Eric Zhongcong and Zhao, Chen and Bansal, Siddhant and Batra, Dhruv and Cartillier, Vincent and Crane, Sean and Do, Tien and Doulaty, Morrie and Erapalli, Akshay and Feichtenhofer, Christoph and Fragomeni, Adriano and Fu, Qichen and Fuegen, Christian and Gebreselasie, Abrham and Gonzalez, Cristina and Hillis, James and Huang, Xuhua and Huang, Yifei and Jia, Wenqi and Khoo, Weslie and Kolar, Jachym and Kottur, Satwik and Kumar, Anurag and Landini, Federico and Li, Chao and Li, Yanghao and Li, Zhenqiang and Mangalam, Karttikeya and Modhugu, Raghava and Munro, Jonathan and Murrell, Tullie and Nishiyasu, Takumi and Price, Will and Puentes, Paola Ruiz and Ramazanova, Merey and Sari, Leda and Somasundaram, Kiran and Southerland, Audrey and Sugano, Yusuke and Tao, Ruijie and Vo, Minh and Wang, Yuchen and Wu, Xindi and Yagi, Takuma and Zhu, Yunyi and Arbelaez, Pablo and Crandall, David and Damen, Dima and Farinella, Giovanni Maria and Ghanem, Bernard and Ithapu, Vamsi Krishna and Jawahar, C. V. and Joo, Hanbyul and Kitani, Kris and Li, Haizhou and Newcombe, Richard and Oliva, Aude and Park, Hyun Soo and Rehg, James M. and Sato, Yoichi and Shi, Jianbo and Shou, Mike Zheng and Torralba, Antonio and Torresani, Lorenzo and Yan, Mingfei and Malik, Jitendra},\n  title     = {Ego4D: Around the {W}orld in 3,000 {H}ours of {E}gocentric {V}ideo},\n  journal   = {CoRR},\n  volume    = {abs/2110.07058},\n  year      = {2021},\n  url       = {https://arxiv.org/abs/2110.07058},\n  eprinttype = {arXiv},\n  eprint    = {2110.07058}\n}\n", "homepage": "", "license": "", "features": {"video_id": {"dtype": "string", "id": null, "_type": "Value"}, "video": {"dtype": "string", "id": null, "_type": "Value"}, "pass_1_narrations": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_1_narrations_timings": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_1_narrations_frames": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_2_narrations": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_2_narrations_timings": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_2_narrations_frames": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_1_summaries": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_1_summaries_timings": {"feature": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_2_summaries": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "pass_2_summaries_timings": {"feature": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "length": -1, "id": null, "_type": "Sequence"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "ego4_d", "config_name": "default", "version": {"version_str": "0.0.0", "description": null, "major": 0, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 260135850, "num_examples": 8584, "dataset_name": "ego4_d"}}, "download_checksums": {}, "download_size": 0, "post_processing_size": null, "dataset_size": 260135850, "size_in_bytes": 260135850}}

ego4d.py

@@ -0,0 +1,136 @@
+# coding=utf-8
+# 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.
+
+# Lint as: python3
+"""A massive-scale, egocentric dataset and benchmark suite collected across 74 worldwide locations and 9 countries, with over 3,670 hours of daily-life activity video."""
+
+
+import os
+import json
+
+import datasets
+
+from .classes import CHARADES_CLASSES
+
+_CITATION = """
+@article{Ego4D2021,
+  author={Grauman, Kristen and Westbury, Andrew and Byrne, Eugene and Chavis, Zachary and Furnari, Antonino and Girdhar, Rohit and Hamburger, Jackson and Jiang, Hao and Liu, Miao and Liu, Xingyu and Martin, Miguel and Nagarajan, Tushar and Radosavovic, Ilija and Ramakrishnan, Santhosh Kumar and Ryan, Fiona and Sharma, Jayant and Wray, Michael and Xu, Mengmeng and Xu, Eric Zhongcong and Zhao, Chen and Bansal, Siddhant and Batra, Dhruv and Cartillier, Vincent and Crane, Sean and Do, Tien and Doulaty, Morrie and Erapalli, Akshay and Feichtenhofer, Christoph and Fragomeni, Adriano and Fu, Qichen and Fuegen, Christian and Gebreselasie, Abrham and Gonzalez, Cristina and Hillis, James and Huang, Xuhua and Huang, Yifei and Jia, Wenqi and Khoo, Weslie and Kolar, Jachym and Kottur, Satwik and Kumar, Anurag and Landini, Federico and Li, Chao and Li, Yanghao and Li, Zhenqiang and Mangalam, Karttikeya and Modhugu, Raghava and Munro, Jonathan and Murrell, Tullie and Nishiyasu, Takumi and Price, Will and Puentes, Paola Ruiz and Ramazanova, Merey and Sari, Leda and Somasundaram, Kiran and Southerland, Audrey and Sugano, Yusuke and Tao, Ruijie and Vo, Minh and Wang, Yuchen and Wu, Xindi and Yagi, Takuma and Zhu, Yunyi and Arbelaez, Pablo and Crandall, David and Damen, Dima and Farinella, Giovanni Maria and Ghanem, Bernard and Ithapu, Vamsi Krishna and Jawahar, C. V. and Joo, Hanbyul and Kitani, Kris and Li, Haizhou and Newcombe, Richard and Oliva, Aude and Park, Hyun Soo and Rehg, James M. and Sato, Yoichi and Shi, Jianbo and Shou, Mike Zheng and Torralba, Antonio and Torresani, Lorenzo and Yan, Mingfei and Malik, Jitendra},
+  title     = {Ego4D: Around the {W}orld in 3,000 {H}ours of {E}gocentric {V}ideo},
+  journal   = {CoRR},
+  volume    = {abs/2110.07058},
+  year      = {2021},
+  url       = {https://arxiv.org/abs/2110.07058},
+  eprinttype = {arXiv},
+  eprint    = {2110.07058}
+}
+"""
+
+_DESCRIPTION = """\
+EGO4D is the world's largest egocentric (first person) video ML dataset and benchmark suite, with 3,600 hrs (and counting) of densely narrated video and a wide range of annotations across five new benchmark tasks. It covers hundreds of scenarios (household, outdoor, workplace, leisure, etc.) of daily life activity captured in-the-wild by 926 unique camera wearers from 74 worldwide locations and 9 different countries. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. The approach to data collection was designed to uphold rigorous privacy and ethics standards with consenting participants and robust de-identification procedures where relevant.
+"""
+
+
+
+class Ego4D(datasets.GeneratorBasedBuilder):
+    """A massive-scale, egocentric dataset and benchmark suite collected across 74 worldwide locations and 9 countries, with over 3,670 hours of daily-life activity video."""
+
+    BUILDER_CONFIGS = [datasets.BuilderConfig(name="default")]
+    DEFAULT_CONFIG_NAME = "default"
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "video_id": datasets.Value("string"),
+                    "video": datasets.Value("string"),
+                    "pass_1_narrations": datasets.features.Sequence(datasets.Value("string")),
+                    "pass_1_narrations_timings": datasets.Sequence(datasets.Value("float32")),
+                    "pass_1_narrations_frames": datasets.features.Sequence(datasets.Value("string")),
+                    "pass_2_narrations": datasets.features.Sequence(datasets.Value("string")),
+                    "pass_2_narrations_timings": datasets.Sequence(datasets.Value("float32")),
+                    "pass_2_narrations_frames": datasets.features.Sequence(datasets.Value("string")),
+                    "pass_1_summaries": datasets.features.Sequence(datasets.Value("string")),
+                    "pass_1_summaries_timings": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
+                    "pass_2_summaries": datasets.features.Sequence(datasets.Value("string")),
+                    "pass_2_summaries_timings": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
+                }
+            ),
+            supervised_keys=None,
+            homepage="",
+            citation=_CITATION,
+        )
+
+    @property
+    def manual_download_instructions(self):
+        return (
+            "To use Ego4D, you must download the data as specified in the instructions on Ego4D website: "
+            "https://ego4d-data.org/docs/start-here/. Please call the dataset loading script with output "
+            "directory used to generate data from CLI like `load_dataset('ego4d', data_dir='/path/to/ego4d_data')` "
+        )
+
+    def _split_generators(self, dl_manager):
+        data_dir = dl_manager.manual_dir
+        videos_dir = os.path.join(data_dir, "v1", "full_scale")
+        narrations = os.path.join(data_dir, "v1", "annotations", "narration.json")
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "videos_folder": videos_dir,
+                    "narrations": narrations,
+                },
+            ),
+        ]
+
+    def _generate_examples(self, videos_folder, narrations):
+        """This function returns the examples."""
+        with open(narrations, "r", encoding="utf-8") as narrations_file:
+            narrations = json.load(narrations_file)
+
+        idx = 0        
+        for video_file in os.listdir(videos_folder):
+            video = os.path.join(videos_folder, video_file)
+            video_id = os.path.splitext(video_file)[0]
+            if video_id not in narrations:
+                continue
+            current_narrations = narrations[video_id]
+            if current_narrations["status"] != "complete":
+                continue
+                
+            pass_1 = current_narrations["narration_pass_1"]        
+            pass_1_narrations = pass_1["narrations"]
+            pass_1_summaries = pass_1["summaries"]
+
+            pass_2 = current_narrations["narration_pass_2"]
+            pass_2_narrations = pass_2["narrations"]
+            pass_2_summaries = pass_2["summaries"]
+            yield idx, {
+                "video_id": video_id,
+                "video": video,
+                "pass_1_narrations": [narration["narration_text"] for narration in pass_1_narrations],
+                "pass_1_narrations_timings": [narration["timestamp_sec"] for narration in pass_1_narrations],
+                "pass_1_narrations_frames": [narration["timestamp_frame"] for narration in pass_1_narrations],
+                "pass_1_summaries": [summary["summary_text"] for summary in pass_1_summaries],
+                "pass_1_summaries_timings": [[summary["start_sec"], summary["end_sec"]] for summary in pass_1_summaries],
+                "pass_2_narrations": [narration["narration_text"] for narration in pass_2_narrations],
+                "pass_2_narrations_timings": [narration["timestamp_sec"] for narration in pass_2_narrations],
+                "pass_2_narrations_frames": [narration["timestamp_frame"] for narration in pass_2_narrations],
+                "pass_2_summaries": [summary["summary_text"] for summary in pass_2_summaries],
+                "pass_2_summaries_timings": [[summary["start_sec"], summary["end_sec"]] for summary in pass_2_summaries],
+
+            }
+
+            idx += 1