onnx-perf-test A simple Python tool to measure the performance of ONNX models. github.com/ibaiGorordo/on…

ONNX-SAM2-Segment-Anything - Updated -Set multimask_output=False on the Pytorch model -> ONNX model will only return one mask per label -Add Bilinear interpolation inside the model for smoother results github.com/ibaiGorordo/ON…

The model still struggles when there are many points per label, but with only few points per object, the results improve

SAM2 (ONNX) - Add box selection - The accuracy is affected by how close the box fits the object - The behaviour with points is not very reliable Code: github.com/ibaiGorordo/ON… Video: youtu.be/gF2Ejq8IKsI

One of the hardest parts of sharing GitHub examples is finding nice images and videos to show what the model can do. I used Flux.1 to generate an image and then ran object detection.

YOLOv9 MIT Object Detection (ONNX) Code: github.com/ibaiGorordo/ON… Video: youtu.be/X_XVkEqgCUM YOLO: github.com/WongKinYiu/YOLO

YOLOv10 Object Detection (ONNX) It is almost the same as my v9 repo but downloads the models from the official repository. Code: github.com/ibaiGorordo/ON… Video: youtu.be/hz9PYZF4ax4 YOLOv10: github.com/THU-MIG/yolov10

NeuflowV2 - Optical Flow (ONNX) "Compared to other state-of-the-art methods, our model achieves a 10x-70x speedup" Added example with Object Detection to mask the detected objects' flow. Code:github.com/ibaiGorordo/ON… Video:youtu.be/S0RnlEHGNrc Neuflow: github.com/neufieldroboti…

本日、tur.ing に入社しました！End-to-End自動運転AIのエンジニアとして挑戦していきます！ Today I joined tur.ing as an engineer working on End-to-End autonomous driving! youtube.com/watch?v=wvZlbA…

入社したばかりですが、チューリング飯再開しました！社員みんな対応できるので、お気軽にどうぞ。一人あたり12,000円で、美味しいご飯と技術トークを楽しみたい方大歓迎！転職意欲は問いません。 少しでも興味あればぜひDMください！

良さそうです。 neuflowのコードでテストしましたが、input_img = input_img / 255.0 の部分（github.com/ibaiGorordo/ON…）をコメントアウトするだけで問題なく動作しました。 SEA-RAFT ONNX Export Colab: colab.research.google.com/drive/1nmgLcAK…

テスラ方式のリアルタイムマルチカメラe2e自動運転出来てきた！

Check out our model in action! Our model’s accuracy feels even more impressive when visualized with Rerun (It would be nice to have flat 3D lines, without using mesh🙏)

Testing Dust3r—here’s the raw output straight from the model. Code coming soon!

Playing with Dust3r—aligning webcam frames to the first one, the raw output: 🎨🎨

Dust3r inference minimal (Pytorch) Minimal & easy-to-use inference code for Dust3r: - Simplified inference code - ONNX export - Limited to 2 images for simplicity Code: github.com/ibaiGorordo/du…

DeDoDe v2 - Feature Detector, Descriptor and Matching (Pytorch) Code: github.com/ibaiGorordo/de… Video:youtu.be/cDHNvORTRjo?si… DeDoDe: github.com/Parskatt/DeDoDe

VGGT Pytorch Inference (WIP) A wrapper around the VGGT model to visualize outputs directly in Rerun. Turn images into point clouds, camera poses, depth maps, and intrinsic easily Code: github.com/ibaiGorordo/vg…