Qualcomm® Innovators Development Kit (QIDK) provides sample applications to demonstrate the capability of Hardware Accelerators for AI, and Software AI stack.
This repository contains sample android applications, which are designed to use components from the following products:
- Qualcomm® Neural Processing SDK for AI Also referred to as SNPE
- Qualcomm® AI Engine Direct SDK Also referred to as QNN
- AI Model Efficiency Tool Kit (AIMET)
- AIMET Model Zoo
Contents of this repository are verified on Snapdragon 8 Gen2, and Snapdragon 8 Gen3 platforms. If users want to try this content on other Qualcomm platforms - please do check with the support e-mail mentioned below.
This Repository is divided into following categories
Qualcomm Wireless Academy has a free course on "AI on Qualcomm Innovators Development Kit".
Course Link : https://qwa.qualcomm.com/course-catalog/AI-on-QIDK
This course is geared toward AI application developers, university students, and AI enthusiasts. This course is applicable, even if a developer is AI SDK on Qualcomm platforms other than QIDK.
All QIDK deliverables are covered in this course in detail with hands-on lab sessions.
Contain examples to use features of above SDKs
| Type | SDK | Details | Link |
|---|---|---|---|
| Model | AI Engine Direct | Model - EnhancementGAN | ReadMe |
| Model | AI Engine Direct | Model - SESR | ReadMe |
Contains examples for :
- Using models, that are not directly supported with AI SDK
- Debug Quantizatin accuracy loss
| Type | SDK | Details | Link |
|---|---|---|---|
| Model Conversion Guide | Neural Processing SDK | Model-Accuracy-Mixed-Precision | ReadMe |
| Model Conversion Guide | Neural Processing SDK | Model-Conversion-Layer-Replacement | ReadMe |
| Model Conversion Guide | Neural Processing SDK | Model-Conversion-UDO-SELU | ReadMe |
Contain end-to-end ready-to-run solutions
| Type | Solution | SDK | API | Model | ReadMe | Demo |
|---|---|---|---|---|---|---|
| NLP | Question Answering | Neural Processing SDK | Native API | Electra-small | ReadMe | Demo |
| NLP | Sentiment Analysis | Neural Processing SDK | Native API | MobileBERT | ReadMe | Demo |
| Vision | Object Detection | Neural Processing SDK | Java API | Mobilenet SSD V2 | ReadMe | Demo |
| Vision | Object Detection YoloNAS | Neural Processing SDK | Native API | YoloNAS | ReadMe | Demo |
| Vision | Image Super Resolution | Neural Processing SDK | Java API | SESR XL | ReadMe | Demo |
| Vision | Image Enhancement | Neural Processing SDK | Java API | EnhancedGAN | ReadMe | Demo |
| Vision | Pose Estimation | Neural Processing SDK | Native API | YoloNAS + HRNet | ReadMe | Demo |
Contain tools to simplify workflow
| Tool | SDK | Details | Link |
|---|---|---|---|
| PySNPE | Neural Processing SDK | Python Interface to SDK tools | ReadMe |
| snpe-docker | Neural Processing SDK | Docker container for SDK | ReadMe |
| snpe-helper | Neural Processing SDK | Python wrapper for C++ API | ReadMe |
All deliverables were periodically verified on latest Qualcomm AI Stack SDK releases. Please report any issues in issues section of GitHub repository.
Pls write to qidk@qti.qualcomm.com for any questions/suggestions
Qualcomm Innovators Development Kit (QIDK) software repository is a project maintained by Qualcomm Innovation Center, Inc.
Please see the LICENSE for more details.