Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Releases: retinify/retinify

retinify v0.2.0

18 Oct 12:18
@SensuiYagi SensuiYagi
v0.2.0
d352beb
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
retinify v0.2.0 Latest
Latest

What's Changed

Read more

Contributors

SensuiYagi
Assets 2
Loading

retinify v0.1.5

06 Sep 11:29
@SensuiYagi SensuiYagi
v0.1.5
0fbacb3

Choose a tag to compare

View all tags
retinify v0.1.5

Highlights

  • Stronger input validation: Pipeline now fixes input pointer types (uint8_t* for 8UC3, float* for 32FC1) and corrects stride validation.
  • New visualization API: Added retinify::ColorizeDisparity(...) (Turbo LUT) to colorize 32-bit disparity maps into 8-bit RGB for quick inspection and demos.
  • Library rename & layout cleanup: Internal directory renamed from core/ → retinify/ and target name unified to retinify for simpler consumption.
  • Docs & samples: README usage snippet clarified, tutorial fixes for Pipeline::Initialize(height, width), and installation docs updated (CUDA 12.x / 13.x, cuDNN 9.x, TensorRT 10.x).

Full Changelog: v0.1.4...v0.1.5

Loading

retinify v0.1.4

19 Aug 04:08
@SensuiYagi SensuiYagi
v0.1.4
1e2aa2d

Choose a tag to compare

View all tags
retinify v0.1.4

Full Changelog: v0.1.3...v0.1.4

Loading

v0.1.3

07 Aug 03:32
@SensuiYagi SensuiYagi
v0.1.3
85f48c2

Choose a tag to compare

View all tags
v0.1.3

retinify v0.1.3

Patch Update
This release focuses on improving documentation clarity and performing a number of internal refactors to streamline the API and benchmarking outputs.

Documentation Improvements

  • Fixed formatting in API reference and index files. (commit 1f38631)
  • Updated latency benchmarks for NVIDIA Jetson Orin Nano in README. (commit c5108de)
  • Clarified latency measurements for each pipeline mode in README. (commit ab9aa1c)
  • Moved “Why Retinify?” section to a more prominent position in README. (commit 852c851)

Refactors

  • Renamed image_dimensions to matching_dimensions in StereoMatchingPipeline. (commit 99125dd)
  • Renamed the resolution enum to model and updated its initialization method for clarity. (commit 9975e70)
  • Removed unused OpenCV ximgproc include and updated logging format in the latency sample. (commit f9443a7)
  • Updated the stereo matching pipeline to use the Mode enum and improved latency benchmarking. (commit 5fcef69)

Full Changelog: v0.1.2...v0.1.3

Loading

v0.1.2

29 Jul 05:50
@SensuiYagi SensuiYagi
v0.1.2
c641a5c

Choose a tag to compare

View all tags
v0.1.2

retinify v0.1.2

🔧 Patch Update
This release fixes several build-configuration issues, refines the project structure, and introduces minor API and documentation improvements.

🔄 Changes

  • 🛠 Build Configuration Fixes

    • Ensured proper inclusion of CUDAToolkit in retinifyConfig.cmake.in and added a status message during configuration. (commit e9b9935) :contentReference[oaicite:0]{index=0}
    • Reordered CUDAToolkit inclusion, adjusted conditional checks, and removed unnecessary toolkit checks in CMakeLists.txt. (commits c82632f, 283306d) :contentReference[oaicite:1]{index=1}
    • Updated CMakeLists.txt to directly include CUDA paths instead of relying on find_package for CUDAToolkit. (commit fc1afec) :contentReference[oaicite:2]{index=2}

  • 🔧 Refactored Include Paths

    • Changed include directives to use the local path.hpp instead of retinify/path.hpp for clearer module boundaries. (commit 9e66b4f) :contentReference[oaicite:3]{index=3}

  • 📦 Project Structure Improvements

    • Added a new 3rdprty directory to better isolate third‑party dependencies and improved overall repository layout. (commit 9f40f69) :contentReference[oaicite:4]{index=4}
    • Updated the Git safe.directory path for googletest submodule initialization to streamline CI workflows. (commit c5fc5fa) :contentReference[oaicite:5]{index=5}

  • 🚀 Pipeline Enhancements

    • Updated the model hash and improved parameter usage logic in the pipeline initialization to ensure consistency with the refreshed weights. (commit 28fae1c) :contentReference[oaicite:6]{index=6}

  • 🆕 API and Documentation Updates

    • Renamed PipelineResolution to Resolution and updated enum values for clarity. (commit 2f095fc) :contentReference[oaicite:7]{index=7}
    • Enhanced docstrings and inline comments for core API functions and enums to improve developer experience. (commit 7e8bc5b) :contentReference[oaicite:8]{index=8}

Full Changelog: v0.1.1...v0.1.2

Loading

v0.1.1

24 Jul 05:00
@SensuiYagi SensuiYagi
v0.1.1
8f1e82e

Choose a tag to compare

View all tags
v0.1.1

retinify v0.1.1 Release Notes

🔧 Patch Update
This release adds a new utility module and refreshes the ONNX model weights (training settings only; minimal change in accuracy).

🔄 Changes

  • 🧰 Added retinify::tools (OpenCV-based Utilities)
    Introduced retinify::tools, a set of utility functions built with OpenCV for easier image manipulation and preprocessing tasks.

  • 🆕 Refreshed ONNX Model Weights
    Updated the ONNX model weights to incorporate only training‑setting adjustments; accuracy improvements are expected to be minor.

Full Changelog: v0.1.0...v0.1.1

Loading

v0.1.0

12 Jul 12:43
@SensuiYagi SensuiYagi
v0.1.0
64b9796

Choose a tag to compare

View all tags
v0.1.0

retinify v0.1.0 Release Notes

🛠 Minor Update
This release aligns the ONNX model with the latest TensorRT update and fixes a critical bug.

🔄 Changes

  • 🎯 Fixed ONNX Model to Float16
    Updated the ONNX model itself to use float16 precision, aligning with the recent TensorRT enhancements.

  • 🐞 Fixed Bug in retinify::Mat
    Resolved an issue in the retinify::Mat class that could lead to incorrect memory access under certain conditions.

Full Changelog: v0.0.2...v0.1.0

Loading

v0.0.2

09 Jul 16:04
@SensuiYagi SensuiYagi
v0.0.2
ce36b4f

Choose a tag to compare

View all tags
v0.0.2

retinify v0.0.2 Release Notes

🛠 Minor Update
This release introduces a performance-focused improvement to the inference pipeline.

🔄 Changes

  • 🌈 Switched Model Input from RGB to Grayscale
    The stereo matching model now takes grayscale images instead of RGB, significantly reducing data transfer and preprocessing costs.

  • ⚡ Improved Inference Speed
    With the grayscale input, average inference latency has been reduced, enabling even faster real-time performance.

📚 See the updated usage notes in the documentation for new input requirements.

Loading

v0.0.1

04 Jul 09:13
@SensuiYagi SensuiYagi
v0.0.1
4e5f6e7

Choose a tag to compare

View all tags
v0.0.1

retinify v0.0.1 Release Notes

🎉 Initial Release
Welcome to the very first public release of retinify! This version lays down the foundation of our real‑time AI stereo vision library.

✨ New Features

  • 🖼️ Real‑time AI Stereo Matching
    High‑precision, real‑time depth estimation powered by AI and GPU acceleration.

  • 🚀 GPU Acceleration
    Builds on TensorRT for maximum throughput.

  • 📦 Easy Build
    Installation made simple with provided build scripts.

  • 📖 Basic Documentation
    Visit our documentation site for setup instructions, usage guides, and more.

Loading