Objects in Context

The Datagen Objects in Context generator allows you to create datasets of behavior-based sequences, solving tasks such as object detection, and object identification

Each sequence is a 10-second recording generated in 15-30 FPS, presenting a human interacting with specific objects in an environment.

The Platform offers control over:

• Human Identity
• Scene – camera location, camera intrinsic, lighting and background
• Object selection
• Distractors selection

For each render you receive full ground truth annotations:

Visual Modalities

• RGB (png)
• Depth (exr)

Textual Modalities

• Camera metadata
• Actor metadata
• Object metadata
• Object key points
• Object bounding cuboids
• Face keypoints
• Face dense key points
• Iris key points
• Pupil key points
• Hands key points
• Body key points