BSD 3-Clause "New" or "Revised" License

技术解读

以下内容由 GitRead 中文社区 编辑团队基于项目 README 与官方资料整理, 结合中文开发者的常见使用场景补充技术选型建议。具体以项目官方文档为准。

Qualcomm AI Hub Models:把「能跑」变成「跑得好」的端侧模型基线

项目背景与痛点

端侧 AI 最耗时间的环节,从来不是写模型结构,而是回答一个问题:这个模型在目标芯片上到底跑多快、占多少内存、掉多少精度?

传统做法是:拿一个 PyTorch 实现,导出 ONNX,用厂商 SDK 转一遍,然后祈祷算子都支持。中间任何一步都可能出问题——某个算子不被 Hexagon NPU 支持,回落到 CPU,延迟直接翻十倍;量化之后精度掉得没法用;预处理和后处理没算进延迟,真机跑起来对不上账。

Qualcomm AI Hub Models 想做的就是把这条链路上的“不确定性”消除掉。它不是又一个模型动物园,而是一份可复现的端侧性能基线:每个模型都给出源码、导出脚本、量化配方、云端真机 profile 结果和预编译资产。你可以在不碰任何 SDK 的情况下,先知道一个模型在 Snapdragon 8 Gen 3 上是 3ms 还是 30ms,再决定要不要投人力。

核心技术揭秘

整个仓库围绕一个约定运转:每个模型是一个独立目录,model.py 是 PyTorch 参考实现,export.py 描述如何导出,test.py 负责真机验证。真正有意思的是中间那层抽象。

Component 拆分是设计上的关键一步。像 Stable Diffusion 这种大型模型,会被拆成 text encoder、UNet、VAE decoder 等若干组件,每个组件通过 get_input_spec() 声明自己的输入形状,独立编译、独立 profile。这解决了三件事:一是大模型无法一次性编译进 NPU 的上下文限制;二是形状固定的要求(QNN 偏好静态 shape,动态维度要枚举成多个变体);三是组件可以跨模型复用。

源码层面的“导出友好”改造是另一个隐性难点。很多上游实现用了 NPU 不友好的算子或控制流,仓库里的 model.py 往往做过等价改写——用受支持的算子组合替掉不支持的,把动态逻辑固化成静态分支。这些改动不明显,但直接决定模型能不能上 HTP。

完整链路大致如下:

graph TD A[Torch 参考实现<br/>models/name/model.py] --> B[组件拆分<br/>Component 与 InputSpec] B --> C[ONNX 导出<br/>静态 shape 变体] C --> D[上传 Qualcomm AI Hub 云端] D --> E[AIMET 量化<br/>int8 / 混合精度 / 低比特权重] E --> F[QNN 编译<br/>HTP / Adreno / CPU] F --> G{"目标运行时"} G -->|"Android"| H[TFLite 加 QNN Delegate] G -->|"Windows 或 Linux"| I[ONNX 加 QNN EP] G -->|"大语言模型"| J[Genie Context Binary] H --> K[真机 Profile<br/>延迟 / 峰值内存 / 精度对齐] I --> K J --> K K --> L[Scorecard 与预编译资产发布]

注意 D 节点:编译和 profile 都在 Qualcomm 的云端真机集群上完成,本地只需要一个 API Token。这是它能把“百来个模型 × 多种芯片 × 多种精度”这套组合矩阵维护下来的根本原因——否则光设备采购就是天文数字。

功能亮点与差异

第一,量化不是事后补丁,而是模型定义的一部分。AIMET 的静态 int8、混合精度、LLM 的低比特权重量化都被写进导出流程,精度损失有明确的对照表。这和多数仓库“给个 float 模型,量化你自己搞”的态度完全不同。

第二,端到端口径。Scorecard 报告的是包含前后处理的整段流水线延迟,而不是单算子累加。YOLO 系列的 NMS、分类模型的 resize 都被纳入考量,这种诚实度在厂商仓库里并不常见。

第三,多运行时覆盖。同一份模型能产出 TFLite(Android)、ONNX(Windows on Snapdragon / Linux)、以及 LLM 场景的 Genie bundle,避免了“Android 一套、PC 一套”的重复建设。

应用场景与落地建议

适合三类场景:选型阶段用它快速筛掉不可行的方案;基线阶段拿官方 scorecard 作为性能下限的参照;移植阶段直接参考它的量化配方和组件拆分方式。

生产落地要注意几点:其一,scorecard 的数字来自参考设备和标准输入,你的预处理若不同(比如换分辨率、改归一化),延迟会变;其二,预编译资产绑定特定 SDK 版本与芯片型号,跨代升级要重新编译;其三,NPU 上下文内存有限,多个模型同时驻留时需要做组件级调度,别指望一把梭;其四,量化后务必在自己的数据集上复测精度,官方给的 ImageNet top-1 不代表你的业务分布。

综合评价

优点:工程完整度极高,从源码到真机数据闭环;量化与编译流程标准化,可复现性强;组件化设计让大模型部署变得可操作;Apache-2.0 许可对企业友好。

缺点:强绑定 Qualcomm 生态,云端编译意味着离线场景和私有模型存在合规顾虑;模型多为参考实现,业务定制仍需二次开发;部分模型的“导出友好”改写牺牲了结构直观性,读源码时容易被绕晕。

一句话:它不是银弹,但它把端侧部署里最脏最累的那段路,替你先走了一遍。

关于

GitHub 原始描述(英文)

Qualcomm® AI Hub Models is our collection of state-of-the-art machine learning models optimized for performance (latency, memory etc.) and ready to deploy on Qualcomm® devices.

描述取自 GitHub 官方 API,保持仓库作者书写的原文。

简要分析

热度分
1721
价值分
820
活跃状态
活跃

语言占比

Dockerfile
Jinja
PowerShell
Python
Shell

Release

README

Qualcomm® AI Hub Models

Release Tag PyPi Python 3.10, 3.11, 3.12, 3.13

The Qualcomm® AI Hub Models are a collection of state-of-the-art machine learning models optimized for deployment on Qualcomm® devices.

See supported: On-Device Runtimes, Hardware Targets & Precision, Chipsets, Devices

 

Install Python Package

Choose the package that fits your use case (both provide the qai-hub-models command-line interface):

# Lightweight CLI for:
# - browsing the catalog (model info, performance, & evals)
# - downloading deployable model assets
pip install qai-hub-models-cli

or

# The full package, for exporting from source, evaluating models, and running end-to-end demos. Includes the CLI.
#
# NOTE for Snapdragon X Elite and Snapdragon X2 Elite users:
# AMDx64 (64-bit) Python is required on Windows.
# Installation will fail when using Windows ARM64 Python.
pip install qai_hub_models

 

Browse and Download Models

Use the CLI to browse the catalog and download deployable assets:

qai-hub-models models # browse the catalog
qai-hub-models info mobilenet_v2  # model details + download options
qai-hub-models numerics mobilenet_v2  # model evaluation results
qai-hub-models perf mobilenet_v2  # model performance results
qai-hub-models fetch mobilenet_v2 --runtime tflite --precision float # download a deployable asset
# ... and more

The CLI also offers a Python API. See the CLI README for full usage instructions.

 

Export from Source and Run A Model on a Physical Device

All models in our directory can be compiled and profiled on a hosted Qualcomm® device.

The downloads and numerics / performance results published with each release are generated by the export and evaluate commands outlined below.

1. Configure AI Hub Workbench Access

Create a Qualcomm® ID, use it to login to Qualcomm® AI Hub Workbench, and configure your API token:

qai-hub configure --api_token API_TOKEN
2. Install Model-Specific Dependencies
qai-hub-models install yolov7
3. Compile, Export, and Evaluate the model.
# Quantize the floating point model to w8a8, convert it to LiteRT format, and profile it on the Samsung Galaxy S25.
qai-hub-models export yolov7 --target-runtime tflite --precision w8a8 --device "Samsung Galaxy S25 (Family)"

# Download a dataset and run evaluation of the LiteRT model on the Samsung Galaxy S25.
qai-hub-models evaluate yolov7 --hub-model-id <Model ID printed by export>

Export and evaluate use Qualcomm® AI Hub Workbench to:

  1. If applicable, Quantize the model (see: Quantization on AI Hub Workbench)
  2. Compile the model for the chosen device and target runtime (see: Compiling Models on AI Hub Workbench).
  3. Download the model.
  4. Profile the compiled model on a real device in the cloud (see: Profiling Models on AI Hub Workbench).
  5. Evaluate the model on a subset of a dataset on a real device in the cloud, and compare on-device model output with PyTorch output (see: Running Inference on AI Hub Workbench)

 

Sample Applications

Native applications that can run our models (with pre and post processing) on physical devices are published in the AI Hub Apps repository.

Python applications are defined for all models (from qai_hub_models.models.<model_name> import App). These apps wrap model inference with pre- and post-processing steps written using torch & numpy. These apps are optimized to be an easy-to-follow example, rather than to minimize prediction time.

 

End-To-End Model Use-Case Demos

Most models have demos that run the model end-to-end (including pre/postprocessing).

1. Install model dependencies.
qai-hub-models install yolov7
2. Demo using Local Inference on CPU / GPU via PyTorch
# Use PyTorch to predict and draw bounding boxes on the provided image.
#
# This runs the model using standard PyTorch kernels via your local CPU or GPU (no NPU support).
qai-hub-models demo yolov7 [--image <path>] [--help]
3. Demo using Accelerated inference on a cloud-hosted Qualcomm Device
# Upload the image and compiled model to a cloud-hosted device,
# run the model, download results, and draw bounding boxes.
#
# This uses a hardware-accelerated target runtime (eg. Lite RT), rather than PyTorch, to run the model.
# Pre and postprocessing may still require numpy and/or PyTorch.
# However, we strive to include as much pre/postprocessing in the accelerated model as is feasible.
#
# NOTE: Not all models have a cloud-hosted demo; some are PyTorch only.
qai-hub-models demo yolov7 [--image ...] --eval-mode on-device --hub-model-id <Model ID printed by export>

End-to-end demos:

  1. Preprocess human-readable input into model input
  2. Run model inference either locally via PyTorch-CPU, or on a cloud-hosted Qualcomm® device.
  3. Postprocess model output to a human-readable format

 

Model Support Data

On-Device Runtimes

Runtime Supported OS
Qualcomm AI Engine Direct Android, Linux, Windows
LiteRT (TensorFlow Lite) Android, Linux
ONNX Android, Linux, Windows

Device Hardware & Precision

Device Compute Unit Supported Precision
CPU FP32, INT16, INT8
GPU FP32, FP16
NPU (includes Hexagon DSP, HTP) FP16*, INT16, INT8

*Some older chipsets do not support fp16 inference on their NPU.

Chipsets

and many more.

Devices

  • Samsung Galaxy S21, S22, S23, S24, and S25 Series
  • Xiaomi 12, 13, 15, and 17
  • Snapdragon X Elite CRD and Snapdragon X2 Elite CRD (Compute Reference Device)
  • Qualcomm RB3 Gen 2, RB5 Gen 2, IQ-8, IQ-9

and many more.

 

Creating a New Model Recipe with Claude

Want to onboard your own model? Follow the instructions in EXTERNAL_ONBOARDING.md to create a new model recipe using Claude.

 

Model Directory

Generative AI

Model README
Image Generation
ControlNet-Canny qai_hub_models.models.controlnet_canny
Stable-Diffusion-v1.5 qai_hub_models.models.stable_diffusion_v1_5
Stable-Diffusion-v2.1 qai_hub_models.models.stable_diffusion_v2_1
Text Generation
Albert-Base-V2-Hf qai_hub_models.models.albert_base_v2_hf
Bert-Base-Uncased-Hf qai_hub_models.models.bert_base_uncased_hf
Distil-Bert-Base-Uncased-Hf qai_hub_models.models.distil_bert_base_uncased_hf
Electra-Bert-Base-Discrim-Google qai_hub_models.models.electra_bert_base_discrim_google
Falcon3-7B-Instruct qai_hub_models.models.falcon_v3_7b_instruct
GPT-OSS-20B qai_hub_models.models.gpt_oss_20b
Gemma-4-E2B-it qai_hub_models.models.gemma_4_e2b_it
Gemma-4-E4B-it qai_hub_models.models.gemma_4_e4b_it
Granite-4.0-Micro qai_hub_models.models.granite_4_0_micro
IBM-Granite-v3.1-8B-Instruct qai_hub_models.models.ibm_granite_v3_1_8b_instruct
IndusQ-1.1B qai_hub_models.models.indus_1b
Intern3.5-VL-2B qai_hub_models.models.intern3_5_vl_2b
Intern3.5-VL-4B qai_hub_models.models.intern3_5_vl_4b
JAIS-6p7b-Chat qai_hub_models.models.jais_6p7b_chat
LFM2.5-1.2B-Instruct qai_hub_models.models.lfm2_5_1_2b_instruct
LFM3-1.7B qai_hub_models.models.lfm3_1_7b
LFM3-VL-2B qai_hub_models.models.lfm3_vl_2b
Llama-SEA-LION-v3.5-8B-R qai_hub_models.models.llama_v3_1_sea_lion_3_5_8b_r
Llama-v3-8B-Instruct qai_hub_models.models.llama_v3_8b_instruct
Llama-v3-ELYZA-JP-8B qai_hub_models.models.llama_v3_elyza_jp_8b
Llama-v3.1-8B-Instruct qai_hub_models.models.llama_v3_1_8b_instruct
Llama-v3.2-1B-Instruct qai_hub_models.models.llama_v3_2_1b_instruct
Llama-v3.2-3B-Instruct qai_hub_models.models.llama_v3_2_3b_instruct
Llama-v3.2-3B-Instruct-SSD qai_hub_models.models.llama_v3_2_3b_instruct_ssd
Llama3-TAIDE-LX-8B-Chat-Alpha1 qai_hub_models.models.llama_v3_taide_8b_chat
Ministral-3-3B-Instruct-2512 qai_hub_models.models.ministral_3_3b_instruct_2512
Mistral-7B-Instruct-v0.3 qai_hub_models.models.mistral_7b_instruct_v0_3
Mobile-Bert-Uncased-Google qai_hub_models.models.mobile_bert_uncased_google
PLaMo-1B qai_hub_models.models.plamo_1b
Phi-3.5-Mini-Instruct qai_hub_models.models.phi_3_5_mini_instruct
Phi-4-Mini-Instruct qai_hub_models.models.phi_4_mini_instruct
Qwen2-7B-Instruct qai_hub_models.models.qwen2_7b_instruct
Qwen2.5-VL-7B-Instruct qai_hub_models.models.qwen2_5_vl_7b_instruct
Qwen3-0.6B qai_hub_models.models.qwen3_0_6b
Qwen3-1.7B qai_hub_models.models.qwen3_1_7b
Qwen3-4B qai_hub_models.models.qwen3_4b
Qwen3-4B-Instruct-2507 qai_hub_models.models.qwen3_4b_instruct_2507
Qwen3-8B qai_hub_models.models.qwen3_8b
Qwen3-VL-2B-Instruct qai_hub_models.models.qwen3_vl_2b_instruct
Qwen3-VL-4B-Instruct qai_hub_models.models.qwen3_vl_4b_instruct
Qwen3-VL-8B-Instruct qai_hub_models.models.qwen3_vl_8b_instruct
Qwen3.5-0.8B qai_hub_models.models.qwen3_5_0_8b
Qwen3.5-2B qai_hub_models.models.qwen3_5_2b
SmolLM2-1.7B-Instruct qai_hub_models.models.smollm2_1_7b_it
SmolVLM2-2.2B-Instruct qai_hub_models.models.smolvlm2_2_2b_instruct

Computer Vision

Model README
Image Classification
Beit qai_hub_models.models.beit
ConvNext-Base qai_hub_models.models.convnext_base
ConvNext-Tiny qai_hub_models.models.convnext_tiny
ConvNextV2-Base qai_hub_models.models.convnext_v2_base
DLA-102-X qai_hub_models.models.dla102x
DenseNet-121 qai_hub_models.models.densenet121
EfficientFormer qai_hub_models.models.efficientformer
EfficientNet-B0 qai_hub_models.models.efficientnet_b0
EfficientNet-B4 qai_hub_models.models.efficientnet_b4
EfficientNet-Lite4 qai_hub_models.models.efficientnet_lite4
EfficientNet-V2-s qai_hub_models.models.efficientnet_v2_s
EfficientViT-b2-cls qai_hub_models.models.efficientvit_b2_cls
EfficientViT-l2-cls qai_hub_models.models.efficientvit_l2_cls
GPUNet qai_hub_models.models.gpunet
GoogLeNet qai_hub_models.models.googlenet
Inception-v3 qai_hub_models.models.inception_v3
Inception-v4 qai_hub_models.models.inception_v4
InternImage qai_hub_models.models.internimage
LeViT qai_hub_models.models.levit
MNASNet05 qai_hub_models.models.mnasnet05
Mobile-VIT qai_hub_models.models.mobile_vit
MobileNet-v2 qai_hub_models.models.mobilenet_v2
MobileNet-v3-Large qai_hub_models.models.mobilenet_v3_large
MobileNet-v3-Small qai_hub_models.models.mobilenet_v3_small
MobileNet-v4 qai_hub_models.models.mobilenet_v4
NASNet qai_hub_models.models.nasnet
NetVLAD qai_hub_models.models.netvlad
RegNet qai_hub_models.models.regnet
RegNet-Y-800MF qai_hub_models.models.regnet_y_800mf
RepViT qai_hub_models.models.repvit
ResNeXt101 qai_hub_models.models.resnext101
ResNeXt50 qai_hub_models.models.resnext50
ResNet101 qai_hub_models.models.resnet101
ResNet18 qai_hub_models.models.resnet18
ResNet50 qai_hub_models.models.resnet50
Sequencer2D qai_hub_models.models.sequencer2d
Shufflenet-v2 qai_hub_models.models.shufflenet_v2
SqueezeNet-1.1 qai_hub_models.models.squeezenet1_1
Swin-Base qai_hub_models.models.swin_base
Swin-Small qai_hub_models.models.swin_small
Swin-Tiny qai_hub_models.models.swin_tiny
SwinV2-Base qai_hub_models.models.swinv2_base
VIT qai_hub_models.models.vit
WideResNet50 qai_hub_models.models.wideresnet50
Image Editing
AOT-GAN qai_hub_models.models.aotgan
DDColor qai_hub_models.models.ddcolor
DnCNN qai_hub_models.models.dncnn
LaMa-Dilated qai_hub_models.models.lama_dilated
NAFNet-DeBlur qai_hub_models.models.nafnet_deblur
NAFNet-DeNoise qai_hub_models.models.nafnet_denoise
Super Resolution
ESRGAN qai_hub_models.models.esrgan
NAFSSR qai_hub_models.models.nafssr
QuickSRNetLarge qai_hub_models.models.quicksrnetlarge
QuickSRNetMedium qai_hub_models.models.quicksrnetmedium
QuickSRNetSmall qai_hub_models.models.quicksrnetsmall
Real-ESRGAN-General-x4v3 qai_hub_models.models.real_esrgan_general_x4v3
Real-ESRGAN-x4plus qai_hub_models.models.real_esrgan_x4plus
SESR-M5 qai_hub_models.models.sesr_m5
SRGAN qai_hub_models.models.srgan
XLSR qai_hub_models.models.xlsr
Semantic Segmentation
DDRNet23-Slim qai_hub_models.models.ddrnet23_slim
DeepLabV3-Plus-MobileNet qai_hub_models.models.deeplabv3_plus_mobilenet
DeepLabXception qai_hub_models.models.deeplab_xception
Detectron2-Panoptic-Segmentation qai_hub_models.models.detectron2_panoptic_seg
EdgeTAM qai_hub_models.models.edgetam
FCN-ResNet50 qai_hub_models.models.fcn_resnet50
FFNet-122NS-LowRes qai_hub_models.models.ffnet_122ns_lowres
FFNet-40S qai_hub_models.models.ffnet_40s
FFNet-54S qai_hub_models.models.ffnet_54s
FFNet-78S qai_hub_models.models.ffnet_78s
FFNet-78S-LowRes qai_hub_models.models.ffnet_78s_lowres
FastSam-S qai_hub_models.models.fastsam_s
FastSam-X qai_hub_models.models.fastsam_x
HRNet-W48-OCR qai_hub_models.models.hrnet_w48_ocr
Mask2Former qai_hub_models.models.mask2former
MaskRCNN qai_hub_models.models.maskrcnn
MediaPipe-Selfie-Segmentation qai_hub_models.models.mediapipe_selfie
MobileSam qai_hub_models.models.mobilesam
PSPNet qai_hub_models.models.pspnet
PidNet qai_hub_models.models.pidnet
PointNet qai_hub_models.models.pointnet
SINet qai_hub_models.models.sinet
SalsaNext qai_hub_models.models.salsanext
Segformer-Base qai_hub_models.models.segformer_base
Segment-Anything-Model-2 qai_hub_models.models.sam2
Segment-Anything-Model-3 qai_hub_models.models.sam3
U2Net-Segmentation qai_hub_models.models.u2net_segmentation
Unet-Segmentation qai_hub_models.models.unet_segmentation
YOLO26-Segmentation qai_hub_models.models.yolo26_seg
YOLOE-Segmentation qai_hub_models.models.yoloe_seg
YOLOv11-Segmentation qai_hub_models.models.yolov11_seg
YOLOv8-Segmentation qai_hub_models.models.yolov8_seg
Video Classification
ResNet-2Plus1D qai_hub_models.models.resnet_2plus1d
ResNet-3D qai_hub_models.models.resnet_3d
ResNet-Mixed-Convolution qai_hub_models.models.resnet_mixed
V-JEPA-2 qai_hub_models.models.vjepa2
Video-MAE qai_hub_models.models.video_mae
Video Generation
First-Order-Motion-Model qai_hub_models.models.fomm
Video Object Tracking
OSNet qai_hub_models.models.osnet
Track-Anything qai_hub_models.models.track_anything
Object Detection
3D-Deep-BOX qai_hub_models.models.deepbox
CavaFace qai_hub_models.models.cavaface
CenterNet-2D qai_hub_models.models.centernet_2d
Conditional-DETR-ResNet50 qai_hub_models.models.conditional_detr_resnet50
D-FINE qai_hub_models.models.dfine
DETR-ResNet101 qai_hub_models.models.detr_resnet101
DETR-ResNet101-DC5 qai_hub_models.models.detr_resnet101_dc5
DETR-ResNet50 qai_hub_models.models.detr_resnet50
DETR-ResNet50-DC5 qai_hub_models.models.detr_resnet50_dc5
Detectron2-Detection qai_hub_models.models.detectron2_detection
Facial-Attribute-Detection qai_hub_models.models.face_attrib_net
HRNetFace qai_hub_models.models.hrnet_face
License-Plates-Detector-Model qai_hub_models.models.lpdm
Lightweight-Face-Detection qai_hub_models.models.face_det_lite
MediaPipe-Face-Detection qai_hub_models.models.mediapipe_face
MediaPipe-Hand-Detection qai_hub_models.models.mediapipe_hand
MediaPipe-Hand-Gesture-Recognition qai_hub_models.models.mediapipe_hand_gesture
MobileFaceNet qai_hub_models.models.mobile_facenet
OWL-V2 qai_hub_models.models.owlv2
OWL-ViT qai_hub_models.models.owl_vit
PPE-Detection qai_hub_models.models.gear_guard_net
Person-Foot-Detection qai_hub_models.models.foot_track_net
RF-DETR qai_hub_models.models.rf_detr
RTMDet qai_hub_models.models.rtmdet
ResNet34-SSD qai_hub_models.models.resnet34_ssd1200
WeDetect qai_hub_models.models.wedetect
YOLO-WORLD qai_hub_models.models.yolo_world
YOLO26-Detection qai_hub_models.models.yolo26_det
YOLOv10-Detection qai_hub_models.models.yolov10_det
YOLOv11-Detection qai_hub_models.models.yolov11_det
YOLOv8-Detection qai_hub_models.models.yolov8_det
YOLOv8-OBB qai_hub_models.models.yolov8_obb
YOLOv9-Detection qai_hub_models.models.yolov9_det
Yolo-R qai_hub_models.models.yolor
Yolo-X qai_hub_models.models.yolox
Yolo-v3 qai_hub_models.models.yolov3
Yolo-v4 qai_hub_models.models.yolov4
Yolo-v5 qai_hub_models.models.yolov5
Yolo-v6 qai_hub_models.models.yolov6
Yolo-v7 qai_hub_models.models.yolov7
YoloV5-Face qai_hub_models.models.yolov5_face
Pose Estimation
AG-Pose qai_hub_models.models.agpose
CenterNet-Pose qai_hub_models.models.centernet_pose
Facial-Landmark-Detection qai_hub_models.models.facemap_3dmm
HRNetPose qai_hub_models.models.hrnet_pose
LiteHRNet qai_hub_models.models.litehrnet
MediaPipe-Pose-Estimation qai_hub_models.models.mediapipe_pose
Movenet qai_hub_models.models.movenet
Posenet-Mobilenet qai_hub_models.models.posenet_mobilenet
RTMPose-Body2d qai_hub_models.models.rtmpose_body2d
SixDRepNet qai_hub_models.models.sixd_repnet
SuperPoint qai_hub_models.models.superpoint
YOLO26-Pose qai_hub_models.models.yolo26_pose
YOLOv11-Pose qai_hub_models.models.yolov11_pose
Gaze Estimation
EyeGaze qai_hub_models.models.eyegaze
Depth Estimation
CREStereo qai_hub_models.models.crestereo
Depth-Anything qai_hub_models.models.depth_anything
Depth-Anything-V2 qai_hub_models.models.depth_anything_v2
Depth-Anything-V3 qai_hub_models.models.depth_anything_v3
Midas-V2 qai_hub_models.models.midas
StereoNet qai_hub_models.models.stereonet
Driver Assistance
BEVDet qai_hub_models.models.bevdet
BEVFormer qai_hub_models.models.bevformer
BEVFusion qai_hub_models.models.bevfusion_det
CVT qai_hub_models.models.cvt
CenterNet-3D qai_hub_models.models.centernet_3d
CenterPoint qai_hub_models.models.centerpoint
GKT qai_hub_models.models.gkt
RangeNet-Plus-Plus qai_hub_models.models.rangenet_plus_plus
StateTransformer qai_hub_models.models.statetransformer
Robotics
ACT qai_hub_models.models.act
HFNet qai_hub_models.models.hfnet

Multimodal

Model README
EasyOCR qai_hub_models.models.easyocr
GR00TN1.5 qai_hub_models.models.grootn15
MiniLM-v2 qai_hub_models.models.minilm_v2
Nomic-Embed-Text qai_hub_models.models.nomic_embed_text
OpenAI-Clip qai_hub_models.models.openai_clip
OpusMT-De-En qai_hub_models.models.opus_mt_de_en
OpusMT-En-De qai_hub_models.models.opus_mt_en_de
OpusMT-En-Es qai_hub_models.models.opus_mt_en_es
OpusMT-En-Fr qai_hub_models.models.opus_mt_en_fr
OpusMT-En-It qai_hub_models.models.opus_mt_en_it
OpusMT-En-Ja qai_hub_models.models.opus_mt_en_ja
OpusMT-En-Ko qai_hub_models.models.opus_mt_en_ko
OpusMT-En-Zh qai_hub_models.models.opus_mt_en_zh
OpusMT-Es-En qai_hub_models.models.opus_mt_es_en
OpusMT-Fr-En qai_hub_models.models.opus_mt_fr_en
OpusMT-It-En qai_hub_models.models.opus_mt_it_en
OpusMT-Ja-En qai_hub_models.models.opus_mt_ja_en
OpusMT-Ko-En qai_hub_models.models.opus_mt_ko_en
OpusMT-Zh-En qai_hub_models.models.opus_mt_zh_en
Pi0.5 qai_hub_models.models.pi05
SigLIP2 qai_hub_models.models.siglip2
TrOCR qai_hub_models.models.trocr

Audio

Model README
Speech Recognition
DeepSpeech2 qai_hub_models.models.deepspeech2
Distil-Whisper qai_hub_models.models.distil_whisper
FunASR-Conformer-EN qai_hub_models.models.funasr_conformer_en
Parakeet-TDT-0.6B-v3 qai_hub_models.models.parakeet_tdt
Wav2Vec2-Conformer-Large-960h qai_hub_models.models.wav2vec2_conformer
Whisper-Base qai_hub_models.models.whisper_base
Whisper-Large-V3-Turbo qai_hub_models.models.whisper_large_v3_turbo
Whisper-Large-V3-Turbo-Quantized qai_hub_models.models.whisper_large_v3_turbo_quantized
Whisper-Medium qai_hub_models.models.whisper_medium
Whisper-Medium-Quantized qai_hub_models.models.whisper_medium_quantized
Whisper-Small qai_hub_models.models.whisper_small
Whisper-Small-Quantized qai_hub_models.models.whisper_small_quantized
Whisper-Tiny qai_hub_models.models.whisper_tiny
Zipformer qai_hub_models.models.zipformer
Audio Classification
YamNet qai_hub_models.models.yamnet
Audio Generation
MeloTTS-EN qai_hub_models.models.melotts_en
MeloTTS-ES qai_hub_models.models.melotts_es
MeloTTS-JA qai_hub_models.models.melotts_ja
MeloTTS-KO qai_hub_models.models.melotts_ko
MeloTTS-ZH qai_hub_models.models.melotts_zh
PiperTTS-DE qai_hub_models.models.pipertts_de
PiperTTS-EN qai_hub_models.models.pipertts_en
PiperTTS-EN-Amy qai_hub_models.models.pipertts_en_amy
PiperTTS-IT qai_hub_models.models.pipertts_it
Voice Activity Detection
Silero-VAD qai_hub_models.models.silero_vad
Speaker Diarization
Pyannote-Speaker-Diarization qai_hub_models.models.pyannote_speaker_diarization

Need help?

Slack: https://aihub.qualcomm.com/community/slack

GitHub Issues: https://github.com/qualcomm/ai-hub-models/issues

Email: [email protected].

LICENSE

Qualcomm® AI Hub Models is licensed under BSD-3. See the LICENSE file.

评论

暂无评论