Version history — VisionMax

Latest stable release

Version 1.2.0

Build date

April 16, 2026· 3 days ago

Size

52.6 MB

What's new

Fixes

Significantly improved overall robustness across capture, rendering, algorithms, tracking, and streaming pipelines
More reliable video startup when KLV packets are missing or incomplete: video frames keep flowing instead of stalling the whole output
Better audio/video continuity on streaming, with RTSP audio monitoring and fallback mechanisms to reduce dropouts

Features

New `image_metadata` pipeline: each image can now be enriched with UTC timestamp, camera identity, sensor type, FOV, platform attitude, and sensor position/orientation
Major STANAG 4609 upgrade with richer ST0601/ST0903 metadata generation and better MISB alignment
Unified input path for gimbal state, navigation, laser telemetry, tracking status, and ground-control events used by the STANAG generator

See what's new in 1.2.0

Download latest version

Version history

Version	Channel	Size	Date	Download
1.2.0	stable	52.6 MB	April 16, 2026	Download
1.1.8	stable	33.0 MB	March 30, 2026	Download
1.0.27	stable	36.5 MB	February 7, 2026	Download
1.0.18	stable	38.5 MB	December 18, 2025	Download

VisionMax changelog

Follow the evolution of VisionMax across each release.

Version 1.2.0

Fixes

Significantly improved overall robustness across capture, rendering, algorithms, tracking, and streaming pipelines
More reliable video startup when KLV packets are missing or incomplete: video frames keep flowing instead of stalling the whole output
Better audio/video continuity on streaming, with RTSP audio monitoring and fallback mechanisms to reduce dropouts
Much more stable video fallback path: clearer emergency test pattern, steadier cadence, and cleaner return to live video
Capture startup now waits for a valid FPGA input format, avoiding “stream started but unusable” situations
Removed the legacy `auto_size=false` path to unify V4L2/FPGA auto-detection and reduce configuration-dependent inconsistencies
More stable overlays and rendering thanks to tighter synchronization between image, detection, stabilization, PiP, and metadata
Fewer unnecessary copies and a more efficient NV12/DMABUF path, reducing stutter risk on overlays, fusion, and fallback rendering
Safer plugin load/unload/restart behavior for both algorithms and render stages, with fewer inconsistent intermediate states
More reliable live streaming control when changing address, port, protocol, codec, and output mode at runtime
Stronger Merio transport, with better handling of large messages and better compatibility with modern status/configuration flows
Explicit legacy compatibility kept for many VisionMAX status names, easing migration from existing control stations
More reliable active gimbal tracking behavior thanks to clearer session handling, cleaner target-loss recovery, and more predictable locking
Safer handling of some piloting commands, especially zoom velocity control and follow-mode transitions

Features

New `image_metadata` pipeline: each image can now be enriched with UTC timestamp, camera identity, sensor type, FOV, platform attitude, and sensor position/orientation
Major STANAG 4609 upgrade with richer ST0601/ST0903 metadata generation and better MISB alignment
Unified input path for gimbal state, navigation, laser telemetry, tracking status, and ground-control events used by the STANAG generator
Added a dedicated `tracking_active` state and richer image metadata to improve consistency between video, tracking, VMTI, and overlays
Richer STANAG overlay able to display configurable timestamp, coordinates, platform/sensor angles, FOV, and target-related information
Stronger AI detection tuned for aerial scenes, more consistent frame-to-frame output
AI tracking enhanced with ReID, helping preserve target identity after occlusion, reframing, or reacquisition
Strengthened and better integrated CSRT tracker, with re-detection logic, health monitoring, and cleaner recovery behavior
New active gimbal tracking modes: follow all trackers, follow trackers from a selected camera, or follow a single tracker
Dedicated services to change follow mode, selected camera ID, and selected tracker ID at runtime
New fusion PiP engine (`FusionAlgo`) for better multi-view overlay composition, with guided placement, HFOV calibration, adjustable opacity, and offset compensation
New configurable PiP path (`PipHintAlgo` + `PipRenderPlugin`) to control PiP position, size, source crop, and transparency
Added automatic “pixel kill” handling to better clean some background areas in fusion compositions
Modular rendering pipeline redesign for cleaner combinations of stabilization, corner boxes, STANAG, PiP, and other overlays
New contract/registry foundation between producers and consumers, making data exchange between algorithms, rendering, tracking, and metadata more reliable
Added a ROS2 `registry_bridge` to expose useful events and values to other modules or integration tools
Extended and unified streaming support with protocol selection (`rtsp`, `rtp`, `udp`, `http`), output mode (`unicast`, `multicast`, `broadcast`), and RTSP transport (`udp`, `tcp`, `all`)
Opus audio added to RTSP streaming with dedicated parameters (buffers, sample rate, channels) for more controlled field operation
Extended runtime streaming configuration: endpoint, protocol, mode, codec, video path, address, bitrate, MTU, and audio/video/KLV queue settings
Stronger MAVLink integration with unified UDP, TCP client/server, or serial transport handling, using a simpler `address` configuration model
Better reporting of VisionMAX/gimbal/camera states toward Merio and MAVLink layers, including streaming, storage, and several system-status items

April 16, 2026

Version 1.1.8

Fixes

Improved overall stability and system handling
Fixed video stream startup when KLV packets are missing by still pushing packets to keep the stream alive
Strengthened Merio transport robustness and handling of large messages
Improved audio/video continuity for streaming and recording
Stabilized fallback test pattern and RTP cadence in fallback mode, with capture startup now waiting for a valid FPGA input format
Removed the legacy `auto_size=false` path to unify V4L2/FPGA auto-detection and avoid inconsistent startup behavior

Features

AI‑reinforced detection better suited to aerial scenes
Improved accuracy and consistency of results
Added two complementary AI trackers: BadasAlgo for responsive tracking, EagleEyesAlgo for robustness and reacquisition
MAVLink added
Merio protocol fundamentally changed approach to simplify usage
Added audio on the RTSP stream, with audio also supported in video recording
Modular refactor and enrichment of the STANAG pipeline with added ST0601/ST0903 tags and MISB-compliant encoding
Added a unified STANAG input path for gimbal state, navigation, laser telemetry, and tracking status
Added a dedicated `tracking_active` state and enriched sensor/platform metadata for ST0601/VMTI generation
Added a timeout on the zoom velocity (`zoom vel`) command
Added an image enhancement plugin for blurry or degraded scenes

March 30, 2026

Version 1.0.27

Fixes

Improved overall stability and system handling

Features

AI‑reinforced detection better suited to aerial scenes
Improved accuracy and consistency of results

February 7, 2026

Version 1.0.17

Fixes

Fixed KLV telemetry overlays and metadata alignment.

Features

Improved tracking performance and stabilization, plus added on-screen display overlay.

December 10, 2025

Version 0.1.16

Fixes

Stabilized the video pipeline to avoid stream drops.
Patched RTSP duration handling to stop unexpected disconnections.
Reduced beacon CPU usage during idle periods.

Features

Updated tracking presets for better day/night behavior.

November 5, 2025