There exist several commonly used datasets in relation to object detection that include COCO (with multiple versions) and ImageNet containing large annotations for 80 and 1000 objects (i.e. classes) respectively. However, very limited datasets are available comprising specific objects identified by visually imapeired people (VIP) such as wheel-bins, trash-Bags, e-Scooters, advertising boards, and bollard. Furthermore, the annotations for these objects are not available in existing sources.

We present below a sample dataset collected using our framework for synthetic data collection that is efficient in terms of time taken to collect and annotate data, and which makes use of free and open source software tools and 3D assets. Our approach provides a large number of systematic variations in synthetic image generation parameters. The approach is highly effective, resulting in a deep learning model with a top-1 accuracy of 72% on the ObjectNet data, which is a new state-of-the-art result.

The MPSC-rPPG dataset comprises photoplethysmograph (rPPG) data with the PPG ground truth, making it a perfect dataset to evaluate various algorithms for extracting PPG, measuring heart rate, heart rate variability from video. The dataset contains facial videos and Blood Volume Pulse (BVP) data captured concurrently.

This dataset is a supplement of the paper DANCE: Domain Adaptation of Networks for Camera Pose Estimation: Learning Camera Pose Estimation Without Pose Labels [1]. The dataset contains a sample scene of a robot garage. The scene was captured by a Leica BLK360 laser scanner, and 16 scans were merged into a single point cloud of 118M colored points. The dataset also contains ~100k synthetically rendered images and scene coordinates generated form the point cloud.