Model Garden provides a collection of state-of-the-art machine learning (ML)
models for vision, text, and audio capabilities. These models are optimized
for use with the Google Tensor SDK, letting you bring powerful AI features
directly to your Pixel devices with seamless on-device performance.

### Depth Estimation

| Model | License |
|---|---|
| [depth_anything](https://github.com/huggingface/transformers/tree/main/src/transformers/models/depth_anything) | Apache-2.0 |
| [depth_anything_v2](https://huggingface.co/depth-anything/Depth-Anything-V2) | Apache-2.0 |
| [midas_v2_1](https://github.com/isl-org/MiDaS) | BSD-3-Clause |

### Face Reconstruction

| Model | License |
|---|---|
| [facemap_3dmm](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/facemap_3dmm/model.py) | BSD-3-Clause |

### Image and text understanding

| Model | License |
|---|---|
| [clip](https://huggingface.co/openai/clip-vit-large-patch14) | MIT |
| [mobileclip_image_encoder](https://github.com/apple/ml-mobileclip) | MIT |
| [mobileclip_text_encoder](https://github.com/apple/ml-mobileclip) | MIT |
| [tinyclip](https://huggingface.co/wkcn/TinyCLIP-ViT-8M-16-Text-3M-YFCC15M) | MIT |
| [trocr_decoder](https://huggingface.co/docs/transformers/en/model_doc/trocr) | MIT |
| [trocr_encoder](https://huggingface.co/docs/transformers/en/model_doc/trocr) | MIT |

### Image Classification

| Model | License |
|---|---|
| [alexnet](https://github.com/pytorch/vision/blob/main/torchvision/models/alexnet.py) | BSD-3-Clause |
| [beit](https://huggingface.co/microsoft/beit-base-patch16-224) | BSD-3-Clause |
| [convnext_base](https://github.com/pytorch/vision/blob/main/torchvision/models/convnext.py) | BSD-3-Clause |
| [convnext_tiny](https://github.com/pytorch/vision/blob/main/torchvision/models/convnext.py) | BSD-3-Clause |
| [densenet121](https://github.com/pytorch/vision/blob/main/torchvision/models/densenet.py) | BSD-3-Clause |
| [efficientformer_l1](https://huggingface.co/timm/efficientformer_l1.snap_dist_in1k) | Apache-2.0 |
| [efficientformerv2_s0](https://huggingface.co/timm/efficientformerv2_s0.snap_dist_in1k) | Apache-2.0 |
| [efficientnet_b0](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnet_b1](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnet_b2](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnet_b3](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnet_b4](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnet_b5](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnet_b6](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnet_b7](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnet_v2_s](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) | BSD-3-Clause |
| [efficientnetv2_m](https://huggingface.co/timm/efficientnetv2_rw_m.agc_in1k) | APACHE-2.0 |
| [efficientvit_cls_b2](https://github.com/mit-han-lab/efficientvit) | BSD-3-Clause |
| [efficientvit_cls_l2](https://github.com/mit-han-lab/efficientvit) | BSD-3-Clause |
| [efficientvit_seg_l2](https://github.com/mit-han-lab/efficientvit) | APACHE-2.0 |
| [googlenet](https://github.com/pytorch/vision/blob/main/torchvision/models/googlenet.py) | BSD-3-Clause |
| [inception_v3](https://github.com/pytorch/vision/blob/main/torchvision/models/inception.py) | BSD-3-Clause |
| [levit](https://huggingface.co/facebook/levit-128S) | APACHE-2.0 |
| [maxvit_t](https://github.com/pytorch/vision/blob/main/torchvision/models/maxvit.py) | BSD-3-Clause |
| [mnasnet0](https://github.com/pytorch/vision/blob/main/torchvision/models/mnasnet.py) | BSD-3-Clause |
| [mobile_vit](https://github.com/apple/ml-cvnets) | BSD-3-Clause |
| [mobilenet_v2 (224_224)](https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv2.py) | APACHE-2.0 |
| [mobilenet_v2 (300_300)](https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv2.py) | APACHE-2.0 |
| [mobilenet_v3_large](https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv3.py) | BSD-3-Clause |
| [mobilenet_v3_small](https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv3.py) | BSD-3-Clause |
| [mobilenetv4_conv_l](https://huggingface.co/timm/mobilenetv4_conv_large.e600_r384_in1k) | BSD-3-Clause |
| [mobilenetv4_conv_m](https://huggingface.co/timm/mobilenetv4_conv_medium.e500_r256_in1k) | BSD-3-Clause |
| [mobilenetv4_conv_s](https://huggingface.co/timm/mobilenetv4_conv_small.e2400_r224_in1k) | BSD-3-Clause |
| [mobilenetv4_hybrid_l](https://huggingface.co/timm/mobilenetv4_hybrid_large.ix_e600_r384_in1k) | BSD-3-Clause |
| [mobilenetv4_hybrid_medium](https://huggingface.co/timm/mobilenetv4_hybrid_medium.ix_e550_r256_in1k) | APACHE-2.0 |
| [mobilesam_vit_t](https://github.com/ChaoningZhang/MobileSAM) | APACHE-2.0 |
| [nasnet](https://huggingface.co/timm/nasnetalarge.tf_in1k) | BSD-3-Clause |
| [nfnet](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/nfnet.py) | BSD-3-Clause |
| [pvt_v2_b1](https://huggingface.co/timm/pvt_v2_b1) | BSD-3-Clause |
| [pvt_v2_b3](https://huggingface.co/timm/pvt_v2_b3) | BSD-3-Clause |
| [regnet](https://github.com/pytorch/vision/blob/main/torchvision/models/regnet.py) | BSD-3-Clause |
| [regnety](https://huggingface.co/qualcomm/RegNet-Y-800MF) | APACHE-2.0 |
| [resnest14d](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/resnest.py) | BSD-3-Clause |
| [resnet101](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [resnet152](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [resnet18](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [resnet50 (224_224)](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [resnet50 (300_300)](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [resnext101](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | AI-HUB-MODELS |
| [resnext50](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [shufflenet_v2](https://github.com/pytorch/vision/blob/main/torchvision/models/shufflenetv2.py) | BSD-3-Clause |
| [squeezenet1 (1_0)](https://github.com/pytorch/vision/blob/main/torchvision/models/squeezenet.py) | BSD-3-Clause |
| [squeezenet1 (1_1)](https://github.com/pytorch/vision/blob/main/torchvision/models/squeezenet.py) | BSD-3-Clause |
| [swin_base](https://github.com/pytorch/vision/blob/main/torchvision/models/swin_transformer.py) | BSD-3-Clause |
| [swin_small](https://github.com/pytorch/vision/blob/main/torchvision/models/swin_transformer.py) | BSD-3-Clause |
| [swin_tiny](https://github.com/pytorch/vision/blob/main/torchvision/models/swin_transformer.py) | BSD-3-Clause |
| [tf_efficientnetv2_m](https://huggingface.co/timm/tf_efficientnetv2_m.in1k) | APACHE-2.0 |
| [vgg16](https://github.com/pytorch/vision/blob/main/torchvision/models/vgg.py) | BSD-3-Clause |
| [vit_base_patch16 (1_3_224_224)](https://github.com/pytorch/vision/blob/main/torchvision/models/vision_transformer.py) | APACHE-2.0 |
| [vit_base_patch16 (224_1_3_224_224)](https://github.com/pytorch/vision/blob/main/torchvision/models/vision_transformer.py) | APACHE-2.0 |
| [vit_small_patch16](https://huggingface.co/pytorch-image-models/blob/main/timm/models/vision_transformer.py) | BSD-3-Clause |
| [wide_resnet101](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [wide_resnet50](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |

### Image Segmentation

| Model | License |
|---|---|
| [hrnet_w48_ocr](https://github.com/HRNet/HRNet-Semantic-Segmentation/tree/HRNet-OCR) | MIT |
| [mediapipe_selfie](https://github.com/google-ai-edge/mediapipe/tree/master/mediapipe/modules/selfie_segmentation) | APACHE-2.0 |
| [unet_segmentation](https://github.com/milesial/Pytorch-UNet) | GPL-3.0 |

### Image Super Resolution

| Model | License |
|---|---|
| [esrgan](https://github.com/xinntao/ESRGAN/) | APACHE-2.0 |

### Object Detection

| Model | License |
|---|---|
| [3d_deep_box](https://github.com/skhadem/3D-BoundingBox/) | MIT |
| [conditional_detr_resnet50](https://huggingface.co/microsoft/conditional-detr-resnet-50) | Apache-2.0 |
| [detr_resnet50](https://huggingface.co/facebook/detr-resnet-50) | Apache-2.0 |
| [detr_resnet50_dc5](https://huggingface.co/facebook/detr-resnet-50-dc5) | Apache-2.0 |
| [detr_resnet101](https://huggingface.co/facebook/detr-resnet-101) | Apache-2.0 |
| [detr_resnet101_dc5](https://huggingface.co/facebook/detr-resnet-101-dc5) | Apache-2.0 |
| [faceattribnet](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/face_attrib_net/model.py) | AI-HUB-MODELS |
| [lightweight_face_detection](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/face_det_lite/model.py) | AI-HUB-MODELS |
| [mediapipe_face_detection](https://github.com/google-ai-edge/mediapipe/tree/master/mediapipe/modules/face_detection) | APACHE-2.0 |
| [mediapipe_hand_detection](https://github.com/zmurez/MediaPipePyTorch) | APACHE-2.0 |
| [person_foot_detection](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/foot_track_net/model.py) | AI-HUB-MODELS |
| [ppe_detection](https://github.com/quic/ai-hub-models/tree/main/qai_hub_models/models/gear_guard_net) | AI-HUB-MODELS |
| [rtdetr_r18vd](https://huggingface.co/PekingU/rtdetr_r18vd) | Apache-2.0 |
| [rtmdet](https://huggingface.co/qualcomm/RTMDet) | Apache-2.0 |
| [yolo_nas_m](https://huggingface.co/hr16/yolo-nas-fp16/blob/main/yolo_nas_m_fp16.onnx) | Apache-2.0 |
| [yolo_nas_s](https://huggingface.co/hr16/yolo-nas-fp16/blob/main/yolo_nas_s_fp16.onnx) | Apache-2.0 |
| [yolo_v4](https://github.com/Tianxiaomo/pytorch-YOLOv4/tree/master) | Apache-2.0 |
| [yolo_v6](https://github.com/meituan/YOLOv6) | GPL-3.0 |
| [yolo_v7](https://github.com/WongKinYiu/yolov7) | GPL-3.0 |
| [yolos_tiny](https://huggingface.co/hustvl/yolos-tiny) | APACHE-2.0 |
| [yolox_tiny](https://github.com/Megvii-BaseDetection/YOLOX) | APACHE-2.0 |

### Optical Flow

| Model | License |
|---|---|
| [raft_small](https://docs.pytorch.org/vision/main/models/generated/torchvision.models.optical_flow.raft_small.html) | BSD-3-Clause |

### Pose Estimation

| Model | License |
|---|---|
| [hrnet_pose](https://github.com/leoxiaobin/deep-high-resolution-net.pytorch) | MIT |
| [lite_hrnet_pose](https://github.com/HRNet/Lite-HRNet) | APACHE-2.0 |
| [mediapipe_pose](https://github.com/zmurez/MediaPipePyTorch) | APACHE-2.0 |
| [movenet](https://github.com/lee-man/movenet-pytorch) | MIT |
| [posenet](https://github.com/michellelychan/posenet-pytorch) | Apache-2.0 |
| [posenet_pose](https://huggingface.co/qualcomm/Posenet-Mobilenet) | Apache-2.0 |
| [rtmpose_body2d](https://huggingface.co/qualcomm/RTMPose-Body2d) | Apache-2.0 |

### Question Answering

| Model | License |
|---|---|
| [tinyroberta](https://huggingface.co/deepset/tinyroberta-squad2) | CC-BY-4.0 |

### Semantic Segmentation

| Model | License |
|---|---|
| [bgnet](https://github.com/thograce/bgnet) | Apache-2.0 |
| [bisenet](https://github.com/ooooverflow/BiSeNet) | No license file |
| [ddrnet23_slim](https://github.com/ydhongHIT/DDRNet) | MIT |
| [deeplabv3_mobilenet_v3_large](https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv3.py) | BSD-3-Clause |
| [deeplabv3_plus_mobilenet](https://github.com/jfzhang95/pytorch-deeplab-xception) | MIT |
| [deeplabv3_resnet101](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [deeplabv3_resnet50](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [fcn_resnet50](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) | BSD-3-Clause |
| [ffnet_122ns_lowres](https://github.com/Qualcomm-AI-research/FFNet) | BSD-3-Clause |
| [ffnet_40s](https://github.com/Qualcomm-AI-research/FFNet) | BSD-3-Clause |
| [ffnet_54s](https://github.com/Qualcomm-AI-research/FFNet) | BSD-3-Clause |
| [ffnet_78s](https://huggingface.co/qualcomm/FFNet-78S) | BSD-3-Clause |
| [ffnet_78s_lowres](https://github.com/Qualcomm-AI-research/FFNet) | BSD-3-Clause |
| [isnet](https://github.com/xuebinqin/DIS) | Apache 2.0 |
| [lraspp_mobilenet_v3_large](https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv3.py) | BSD-3-Clause |
| [pidnet](https://github.com/XuJiacong/PIDNet) | MIT |
| [sam_vit_b](https://github.com/facebookresearch/segment-anything) | APACHE-2.0 |
| [sam_vit_l](https://github.com/facebookresearch/segment-anything) | APACHE-2.0 |
| [segformer](https://huggingface.co/nvidia/segformer-b0-finetuned-ade-512-512) | NVIDIA-SCSL |
| [segment_anything_model](https://huggingface.co/facebook/sam-vit-base) | Apache-2.0 |
| [u2net_full](https://github.com/xuebinqin/U-2-Net) | APACHE-2.0 |
| [u2net_lite](https://github.com/xuebinqin/U-2-Net) | APACHE-2.0 |

### Speech Recognition

| Model | License |
|---|---|
| [conformer](https://huggingface.co/speechbrain/asr-conformer-loquacious) | BSD-2-Clause |
| [deepspeech](https://github.com/pytorch/audio) | BSD-2-Clause |
| [huggingface_wavlm_base_plus](https://huggingface.co/microsoft/wavlm-base-plus) | Apache-2.0 |
| [parakeet_ctc](https://huggingface.co/nvidia/parakeet-ctc-1.1b) | Apache-2.0 |
| [torchaudio_emformer_rnnt_base](https://github.com/pytorch/audio) | BSD-2-Clause |
| [wav2vec2_base_960h](https://huggingface.co/facebook/wav2vec2-base-960h) | APACHE-2.0 |
| [whisper_base](https://github.com/openai/whisper/tree/main) | MIT |
| [whisper_medium](https://github.com/openai/whisper/tree/main) | MIT |
| [whisper_large_v2](https://github.com/huggingface/transformers/tree/v4.42.3/src/transformers/models/whisper) | MIT |
| [whisper_small](https://github.com/openai/whisper/tree/main) | MIT |
| [whisper_tiny](https://github.com/openai/whisper/tree/main) | MIT |

### Super Resolution

| Model | License |
|---|---|
| [quicksrnet_large](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/quicksrnet) | BSD-3-Clause |
| [quicksrnet_medium](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/quicksrnet) | BSD-3-Clause |
| [quicksrnet_small](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/quicksrnet) | BSD-3-Clause |
| [real_esrgan_general_x4v3](https://github.com/xinntao/Real-ESRGAN) | BSD-3-Clause |
| [real_esrgan_x4plus](https://github.com/xinntao/Real-ESRGAN) | BSD-3-Clause |
| [xlsr](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/xlsr) | BSD-3-Clause |
| [sesr_m5](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/sesr) | BSD-3-Clause |

### Text Classification

| Model | License |
|---|---|
| [distilbert](https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english) | Apache-2.0 |
| [mobilebert](https://huggingface.co/lordtt13/emo-mobilebert) | Apache-2.0 |

### Small Language Models (SLMs)

| Model | License |
|---|---|
| [EmbeddingGemma 2 (740M)](https://huggingface.co/litert-community/embeddinggemma-2-740m-litert-lm) | [Gemma Terms of Use](https://ai.google.dev/gemma/terms) |
| [EmbeddingGemma (300M)](https://huggingface.co/litert-community/embeddinggemma-300m) | [Gemma Terms of Use](https://ai.google.dev/gemma/terms) |
| [Gemma 3 (270M)](https://huggingface.co/google/gemma-3-270m) | [Gemma Terms of Use](https://ai.google.dev/gemma/terms) |
| [Gemma 3 (1B)](https://huggingface.co/litert-community/Gemma3-1B-IT) | [Gemma Terms of Use](https://ai.google.dev/gemma/terms) |
| [FunctionGemma (270M) FT Mobile Actions](https://huggingface.co/litert-community/functiongemma-270m-ft-mobile-actions) | [Gemma Terms of Use](https://ai.google.dev/gemma/terms) |