Model Garden fournit une collection de modèles de machine learning (ML) de pointe pour les fonctionnalités de vision, de texte et d'audio. Ces modèles sont optimisés pour être utilisés avec le SDK Google Tensor, ce qui vous permet d'intégrer de puissantes fonctionnalités d'IA directement à vos appareils Pixel, avec des performances fluides sur l'appareil.
Estimation de la profondeur
| Modèle | Licence |
|---|---|
| depth_anything_v2 | Apache-2.0 |
| midas_v2_1 | BSD-3-Clause |
Reconstruction du visage
| Modèle | Licence |
|---|---|
| facemap_3dmm | BSD-3-Clause |
Compréhension des images et du texte
| Modèle | Licence |
|---|---|
| clip | MIT |
| mobileclip_image_encoder | MIT |
| mobileclip_text_encoder | MIT |
| tinyclip | MIT |
Classification d'images
| Modèle | Licence |
|---|---|
| alexnet | BSD-3-Clause |
| beit | BSD-3-Clause |
| convnext_base | BSD-3-Clause |
| convnext_tiny | BSD-3-Clause |
| densenet121 | BSD-3-Clause |
| efficientformer_l1 | Apache-2.0 |
| efficientformerv2_s0 | Apache-2.0 |
| efficientnet_b0 | BSD-3-Clause |
| efficientnet_b1 | BSD-3-Clause |
| efficientnet_b2 | BSD-3-Clause |
| efficientnet_b3 | BSD-3-Clause |
| efficientnet_b4 | BSD-3-Clause |
| efficientnet_b5 | BSD-3-Clause |
| efficientnet_b6 | BSD-3-Clause |
| efficientnet_b7 | BSD-3-Clause |
| efficientnet_v2_s | BSD-3-Clause |
| efficientnetv2_m | APACHE-2.0 |
| efficientvit_cls_b2 | BSD-3-Clause |
| efficientvit_cls_l2 | BSD-3-Clause |
| efficientvit_seg_l2 | APACHE-2.0 |
| googlenet | BSD-3-Clause |
| inception_v3 | BSD-3-Clause |
| levit | APACHE-2.0 |
| maxvit_t | BSD-3-Clause |
| mnasnet0 | BSD-3-Clause |
| mobile_vit | BSD-3-Clause |
| mobilenet_v2 | APACHE-2.0 |
| mobilenet_v3_large | BSD-3-Clause |
| mobilenet_v3_small | BSD-3-Clause |
| mobilenetv4_conv_l | BSD-3-Clause |
| mobilenetv4_conv_m | BSD-3-Clause |
| mobilenetv4_conv_s | BSD-3-Clause |
| mobilenetv4_hybrid_l | BSD-3-Clause |
| mobilenetv4_hybrid_medium | APACHE-2.0 |
| nfnet | BSD-3-Clause |
| pvt_v2_b1 | BSD-3-Clause |
| pvt_v2_b3 | BSD-3-Clause |
| regnety | APACHE-2.0 |
| resnest14d | BSD-3-Clause |
| resnet101 | BSD-3-Clause |
| resnet152 | BSD-3-Clause |
| resnet18 | BSD-3-Clause |
| resnet50 | BSD-3-Clause |
| resnext101 | AI-HUB-MODELS |
| resnext50 | BSD-3-Clause |
| shufflenet_v2 | BSD-3-Clause |
| squeezenet1 | BSD-3-Clause |
| swin_small | BSD-3-Clause |
| swin_tiny | BSD-3-Clause |
| tf_efficientnetv2_m | APACHE-2.0 |
| vgg16 | BSD-3-Clause |
| vit_base_patch16 | APACHE-2.0 |
| vit_small_patch16 | BSD-3-Clause |
| wide_resnet101 | BSD-3-Clause |
| wide_resnet50 | BSD-3-Clause |
Segmentation d'images
| Modèle | Licence |
|---|---|
| hrnet_w48_ocr | MIT |
| mediapipe_selfie | APACHE-2.0 |
| unet_segmentation | GPL-3.0 |
Super-résolution d'images
| Modèle | Licence |
|---|---|
| esrgan | APACHE-2.0 |
Détection d'objet
| Modèle | Licence |
|---|---|
| 3d_deep_box | MIT |
| conditional_detr_resnet50 | Apache-2.0 |
| detr_resnet50 | Apache-2.0 |
| detr_resnet50_dc5 | Apache-2.0 |
| detr_resnet101 | Apache-2.0 |
| detr_resnet101_dc5 | Apache-2.0 |
| faceattribnet | AI-HUB-MODELS |
| lightweight_face_detection | AI-HUB-MODELS |
| mediapipe_hand_detection | APACHE-2.0 |
| person_foot_detection | AI-HUB-MODELS |
| ppe_detection | AI-HUB-MODELS |
| yolo_v4 | Apache-2.0 |
| yolo_v6 | GPL-3.0 |
| yolo_v7 | GPL-3.0 |
| yolos_tiny | APACHE-2.0 |
| yolox_tiny | APACHE-2.0 |
Estimation des postures
| Modèle | Licence |
|---|---|
| hrnet_pose | MIT |
| lite_hrnet_pose | APACHE-2.0 |
| mediapipe_pose | APACHE-2.0 |
| movenet | MIT |
Systèmes de questions-réponses
| Modèle | Licence |
|---|---|
| tinyroberta | CC-BY-4.0 |
Segmentation sémantique
| Modèle | Licence |
|---|---|
| bgnet | Apache-2.0 |
| bisenet | Aucun fichier de licence |
| ddrnet23_slim | MIT |
| deeplabv3_mobilenet_v3_large | BSD-3-Clause |
| deeplabv3_plus_mobilenet | MIT |
| deeplabv3_resnet101 | BSD-3-Clause |
| deeplabv3_resnet50 | BSD-3-Clause |
| fcn_resnet50 | BSD-3-Clause |
| ffnet_122ns_lowres | BSD-3-Clause |
| ffnet_40s | BSD-3-Clause |
| ffnet_54s | BSD-3-Clause |
| ffnet_78s_lowres | BSD-3-Clause |
| isnet | Apache 2.0 |
| lraspp_mobilenet_v3_large | BSD-3-Clause |
| sam_vit_b | APACHE-2.0 |
| sam_vit_l | APACHE-2.0 |
| segformer | NVIDIA-SCSL |
| segment_anything_model | Apache-2.0 |
| u2net_full | APACHE-2.0 |
| u2net_lite | APACHE-2.0 |
Reconnaissance vocale
| Modèle | Licence |
|---|---|
| deepspeech | BSD-2-Clause |
| torchaudio_emformer_rnnt_base | BSD-2-Clause |
| wav2vec2_base_960h | APACHE-2.0 |
Super-résolution
| Modèle | Licence |
|---|---|
| quicksrnet_large | BSD-3-Clause |
| quicksrnet_small | BSD-3-Clause |
| real_esrgan_general_x4v3 | BSD-3-Clause |
| real_esrgan_x4plus | BSD-3-Clause |
| xlsr | BSD-3-Clause |
Classification de texte
| Modèle | Licence |
|---|---|
| distilbert | Apache-2.0 |
| mobilebert | Apache-2.0 |