ailia-ai/ailia-models
The collection of pre-trained, state-of-the-art AI models for ailia SDK
2.3kStars
361Forks
49Claude Commits
PythonLanguage
action-recognitionanomaly-detectionaudio-processingbackground-removalcrowd-countingdeep-learningembeddingsface-detectionface-recognitionfashion-aiganhand-detectionimage-classificationimage-segmentationllmneural-networkobject-detectionobject-recognitionobject-trackingpose-estimation
First Claude commit: Feb 27, 2026Last Claude commit: 2mo agoDiscovered: Mar 22, 2026
Recent Claude Commits
Remove --fp16 option from qwen2_vl.py
f2c60372mo agoauthor_emailUse fp32 model as quantization source, remove fp16 conversion
14152be2mo agoauthor_emailUse opt model as quantization source instead of fp16
aa6f5542mo agoauthor_emailRevert external data format and disk cleanup optimizations
22b5fa92mo agoauthor_emailFix export to use external data format, restore PB_PATH for int4
0c327a42mo agoauthor_emailRevert "Support 2B and 7B int4 model export with --model argument"
ca6858f2mo agoauthor_emailSupport 2B and 7B int4 model export with --model argument
cb0aed62mo agoauthor_emailPack int4 model into single ONNX file, add version info to export script
8eecb972mo agoauthor_emailFix int4 export: use fp32 base, handle fp16->fp32 conversion, external data
d01fb462mo agoauthor_emailAdd Qwen2-VL-2B int4 quantized model support using Microsoft Olive
67853cb2mo agoauthor_emailSort models by publication date within each category table
9d665742mo agoauthor_emailAdd publication dates to all model tables in README.md
c6db48d2mo agoauthor_emailUpdate output.png with inverse depth visualization
4447eae3mo agoauthor_emailRevert preprocess change and remove compare script
41c16b03mo agoauthor_emailFix depth visualization: convert metric depth to inverse depth
536e3683mo agoauthor_emailFix preprocessing order: resize on uint8 before float conversion
9cc7e493mo agoauthor_emailAdd Torch vs ONNX comparison script for depth_anything_v3
25ef6053mo agoauthor_emailRemove unused export directory for DepthPro
1dc814b3mo agoauthor_emailAdd HuggingFace ONNX reference URL to DepthPro README
a6b4c6a3mo agoauthor_emailSwitch DepthPro to FP16 single-file ONNX (1.8GB) using HF Transformers version
0de4ef43mo agoauthor_emailAdd DepthPro (Apple ml-depth-pro) to depth estimation models
bebaa4a3mo agoauthor_emailRevert inference script changes for Depth Anything V2
1318e723mo agoauthor_emailFix Depth Anything V2 export and inference for correct ONNX output
e332a1e3mo agoauthor_emailFix DA3 model loading and preprocessing for correct depth output
29a5aa33mo agoauthor_emailAdd Depth Anything V3 to README.md and download_all_models.sh
52b343e3mo agoauthor_emailFix DA3 export for ONNX compatibility and update inference script
c26cbd93mo agoauthor_emailImprove DA3 export script with ONNX wrapper and HuggingFace download
06c00683mo agoauthor_emailAdd Depth Anything V3 support with LICENSE (Apache 2.0)
9b5ec463mo agoauthor_emailAdd LICENSE file for depth_anything_v2 (Apache 2.0)
7bce3053mo agoauthor_emailUpdate depth_anything_v2 required ailia SDK version to 1.2.16
1c559b03mo agoauthor_emailUpdate depth_anything_v2 model paths and add Netron links
0bee3603mo agoauthor_emailFix Depth Anything V2: adjust output shape and remove vitg
853cbeb3mo agoauthor_emailAdd Depth Anything V2 model support with ONNX export
2d50a2b3mo agoauthor_emailRemove duplicated text folder from v2-pro (now shared from v2)
6da1ef73mo agoauthor_emailUpdate v2-pro to import text module from v2 instead of local copy
1e9c1fd3mo agoauthor_emailAdd gpt-sovits-v2-pro to README.md and download_all_models.sh
c1719203mo agoauthor_emailFix v2Pro ONNX export: add sv_emb support and real weight conversion
97e67c13mo agoauthor_emailAdd GPT-SoVITS v2pro ONNX conversion and inference support
a8495523mo agoauthor_emailMatch official BEVFormer/nuScenes Box.render() drawing convention
b2ac61f3mo agoauthor_emailMatch nuScenes Box.corners() convention for 3D bbox rendering
9324b673mo agoauthor_emailAdd *.pth to .gitignore for pretrained checkpoint files
dac709b3mo agoauthor_emailUse nuScenes mini dataset images with exact calibration and fix spatial cross-attention
24666253mo agoauthor_emailLoad pretrained weights and add camera-aware cross-attention for BEVFormer ONNX
afcf3f53mo agoauthor_emailAdd 3D bbox projection onto camera images and limit max detections
98114a23mo agoauthor_emailMove ONNX export code to export/ subdirectory
a4530c83mo agoauthor_emailUse 6 surround-view camera images for BEVFormer demo input
a6b5b013mo agoauthor_emailMerge .onnx.data into single file and download ONNX from GCS
243fd513mo agoauthor_emailReplace Deformable Attention with grid_sample for full ONNX export
056e5ef3mo agoauthor_emailAdd BEVFormer (BEV 3D object detection) to object_detection_3d
21937b23mo agoauthor_email