Datasets:

psi-benchmark
/

PSI-Benchmark

@@ -20,8 +20,8 @@ PSI-2.0 covers 196 scenes including 110 scenes from PSI-1.0 and contains the lab
 These innovative labels can enable computer vision tasks like pedestrian intent/behavior prediction, vehicle-pedestrian interaction segmentation, and video-to-language mapping for explainable algorithms. The dataset also contains driving dynamics and driving decision-making reasoning explanations.
-![image](./images/dataset_multimodal.jpg)
-![image](./images/dataset_temporal.jpg)
 ## Supported Tasks
 - Pedestrian crossing intent prediction and reasoning generation
@@ -32,8 +32,28 @@ These innovative labels can enable computer vision tasks like pedestrian intent/
 - Visual-language matching and causal reasoning
 - General classification, object detection, segmentation, and more tasks.
 ## Bias, Privacy, Social Impact, and Limitations Considerations
-- **Bias and Representativeness**: Our 74 annotators (44 males, 30 females) range in age from 19 to 77, with equal distribution across three age groups (19–30, 31–54, $>=$55), and all hold valid U.S. driver’s licenses. Driving experience varies: ~25\% drive ~5,000 miles/year, ~40\% drive ~10,000 miles/year, and the remainder >15,000 miles/year. This diversity supports a broad range of perspectives in pedestrian intent estimation and reasoning. However, we acknowledge that the dataset is U.S.-centric, with driving scenes collected from multiple U.S. cities, and may not generalize globally.
 - **Privacy and Consent**: All video and annotation data were reviewed and approved by the Indiana University legal department under an IRB-approved protocol. Dataset users must agree to use it solely for research purposes. While the current release does not include face or license plate blurring, we acknowledge this limitation and plan to apply de-identification and anonymization protocols in future updates. Although efforts were made to anonymize the data, users should be cautious about potential privacy concerns, particularly if faces or license plates are visible in the footage. Ensure compliance with local data protection regulations when using or sharing the dataset.
 - **Social Impact**: We recognize the risk of overfitting models to subjective human reasoning. To address this, PSI includes annotations from diverse annotators and preserves disagreement across labels. This enables the development of models that reason under uncertainty rather than relying solely on consensus.
 - **Limitations**: This dataset offers valuable insights into driver-pedestrian interactions and reasoning for autonomous driving research, but users should be aware of a few limitations: (1) Scale: the dataset is smaller than some large-scale benchmarks, which may affect generalization in data-hungry models. (2) Geographic scope: All videos were collected in U.S. cities, which may limit applicability to regions with different traffic rules, infrastructure, or pedestrian behavior. (3) Demographic representation: The dataset may not comprehensively represent diverse pedestrian demographics (e.g., age, mobility, clothing styles), which could lead to biased behavior prediction or reasoning outcomes. (4) Annotation Subjectivity: If the dataset includes reasoning or intention annotations, these may be subject to annotator bias or interpretation variability, especially in ambiguous scenarios. (5) The videos may not reflect the latest advancements in camera technology, changes in traffic behavior, vehicle technologies, or pdestrian norms, which could affect the relevance of the data for current autonomous driving systems.

 These innovative labels can enable computer vision tasks like pedestrian intent/behavior prediction, vehicle-pedestrian interaction segmentation, and video-to-language mapping for explainable algorithms. The dataset also contains driving dynamics and driving decision-making reasoning explanations.
+![image](./dataset_multimodal.jpg)
+![image](./dataset_temporal.jpg)
 ## Supported Tasks
 - Pedestrian crossing intent prediction and reasoning generation
 - Visual-language matching and causal reasoning
 - General classification, object detection, segmentation, and more tasks.
+## Dataset Preparation
+Download the PSI 2.0 Dataset videos and annotations from Hugging Face or [[PSI Homepage](http://pedestriandataset.situated-intent.net)]. Move *\*.zip* files to the dataset *ROOT_PATH*, and unzip them by
+```shell
+    cd ROOT_PATH # e.g., root/Dataset
+    unzip '*.zip' -d .
+    rm *.zip
+```
+The extracted folder contains all videos (Train/Val):
+-  *ROOT_PATH/PSI_Videos/videos*.
+The extracted folder contains all annotations of the PSI 2.0 Dataset (Train/Val)
+- *ROOT_PATH/PSI2.0_TrainVal/annotations/cognitive_annotation_key_frame*
+- *ROOT_PATH/PSI2.0_TrainVal/annotations/cv_annotation*
+and the train/val/test splits:
+- *ROOT_PATH/PSI2.0_TrainVal/splits/PSI2_split.json*.
 ## Bias, Privacy, Social Impact, and Limitations Considerations
+- **Bias and Representativeness**: Our 74 annotators (44 males, 30 females) range in age from 19 to 77, with equal distribution across three age groups (19–30, 31–54, >=$55), and all hold valid U.S. driver’s licenses. Driving experience varies: ~25\% drive ~5,000 miles/year, ~40\% drive ~10,000 miles/year, and the remainder >15,000 miles/year. This diversity supports a broad range of perspectives in pedestrian intent estimation and reasoning. However, we acknowledge that the dataset is U.S.-centric, with driving scenes collected from multiple U.S. cities, and may not generalize globally.
 - **Privacy and Consent**: All video and annotation data were reviewed and approved by the Indiana University legal department under an IRB-approved protocol. Dataset users must agree to use it solely for research purposes. While the current release does not include face or license plate blurring, we acknowledge this limitation and plan to apply de-identification and anonymization protocols in future updates. Although efforts were made to anonymize the data, users should be cautious about potential privacy concerns, particularly if faces or license plates are visible in the footage. Ensure compliance with local data protection regulations when using or sharing the dataset.
 - **Social Impact**: We recognize the risk of overfitting models to subjective human reasoning. To address this, PSI includes annotations from diverse annotators and preserves disagreement across labels. This enables the development of models that reason under uncertainty rather than relying solely on consensus.
 - **Limitations**: This dataset offers valuable insights into driver-pedestrian interactions and reasoning for autonomous driving research, but users should be aware of a few limitations: (1) Scale: the dataset is smaller than some large-scale benchmarks, which may affect generalization in data-hungry models. (2) Geographic scope: All videos were collected in U.S. cities, which may limit applicability to regions with different traffic rules, infrastructure, or pedestrian behavior. (3) Demographic representation: The dataset may not comprehensively represent diverse pedestrian demographics (e.g., age, mobility, clothing styles), which could lead to biased behavior prediction or reasoning outcomes. (4) Annotation Subjectivity: If the dataset includes reasoning or intention annotations, these may be subject to annotator bias or interpretation variability, especially in ambiguous scenarios. (5) The videos may not reflect the latest advancements in camera technology, changes in traffic behavior, vehicle technologies, or pdestrian norms, which could affect the relevance of the data for current autonomous driving systems.