GStreamer Expands ML Inference For Media
Collabora announced GStreamer 1.28, expanding machine-learning and AI inference capabilities for media pipelines. The release adds improved ONNX Runtime support, a LiteRT (formerly TensorFlow Lite) plugin, ModelInfo sidecar files, broader tensor decoding (including YOLO v8–v11 and FastSAM), and NPU backends for STM32MP2, Amlogic A311D, and NXP i.MX 8M Plus. These changes simplify pipeline creation, metadata handling, overlays, and batching on embedded Linux and Android-class systems.
Key Points
- 1Introduces new inference engines, tensor decoders, and ModelInfo sidecar for stable tensor handling
- 2Enables NPU acceleration on embedded platforms (STM32MP2, Amlogic A311D, i.MX 8M Plus)
- 3Simplifies pipeline creation with tensordecodebin, improved metadata handling, overlays, and batching
Scoring Rationale
Official, practical release with broad embedded applicability; scope focused mainly on media pipelines and embedded systems.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
