On this page, you can find practical examples, how-tos, publications, and other non-formal documentation links that can help you dive into Savant.

• ML/AI Examples
• Utility And Coding Examples

To help novice users to dive into Savant, we have prepared several ready-to-use examples, which one can download, build and launch. In addition, every sample includes all you need to run it: the source code, models, build file, sample data, and a short README.md which covers the purpose and gives a brief explanation. Some samples are also accompanied by the how-to guides published on the Medium portal, where one can study them step-by-step.

Sample location: peoplenet_detector

Sample location: traffic_meter

Sample location: intersection_traffic_meter

Sample location fisheye_line_crossing

Sample location: area_object_counting

Sample location: age gender recognition

Sample location: yolov11_seg

Sample location: face_reid

Sample location: LPR

Sample location: keypoint_detection

Sample location: super_resolution

Sample location: panoptic_driving_perception

The sample shows how to synchronize multiple RTSP streams by RTCP SR NTP data with the Savant Retina RTSP adapter.

Sample location: retina_rtsp_rtcp_sr

Sample location: opencv_cuda_bg_remover_mog2

Sample location: animegan

Sample location: conditional_video_processing

A simple pipeline demonstrates how multiplexed processing works in Savant. In the demo, two RTSP streams are ingested in the module and processed with the PeopleNet model.

Sample location: multiple_rtsp

A pipeline shows how to send a video file for processing and get resulting video file after processing.

Sample location: file_processing

A very primitive source-sink pipeline testing that RTSP cam normally processed by NVDEC and thus, is compatible with Savant.

Sample location: rtsp_cam_compatibility_test

A simple pipeline demonstrates how GigE Vision Source Adapter works in Savant. In the demo video from one GigE Vision camera is passed as raw-rgba frames, and another one is passed as HEVC-encoded frames. Both streams are passed to Always-On-RTSP sinks.

Sample location: multiple_gige

A demo showing how to inject ground truth metadata into frames and use them to estimate the model performance.

Sample location: source_adapter_with_json_metadata

A sample demonstrating the use of OpenTelemetry in Savant.

Sample location: telemetry

A sample demonstrating the use of Kafka-Redis adapters in Savant.

Sample location: kafka_redis_adapter

A sample demonstrating the use of USB camera producing MJPEG in Savant.

Sample location: mjpeg_usb_cam

A sample demonstrating the use of pass-through mode in Savant.

Sample location: pass_through_processing

A sample demonstrating the use of RT-DETR model in Savant.

Sample location: rtdetr

A pipeline demonstrating the use of Auxiliary Streams in Savant.

Sample location: auxiliary_streams

A pipeline demonstrates processing of streams at the original resolution, i.e. without scaling to a single resolution.

Sample location: original_resolution_processing

A pipeline demonstrates the router service distributing video streams between screenshot maker and video archiving sink.

Sample location: router

A pipeline demonstrating the Meta Merge service for merging metadata from multiple parallel inference pipelines. The sample splits the frame into left and right ROIs, runs YOLO11 on each half in separate module instances, merges objects using export_complete_object_trees and import_object_trees, and visualizes the combined results.

Sample location: meta_merge

The sample shows how to use the module with uridecodebin input to feed downstream elements with videostreams.

Sample location: uridecodebin