Product Overview
Equipped with the Di1 SoM, this product is the optimal solution for enabling customers to immediately begin evaluation of the Di1 SoM and Di1, as well as application development.
Product Specifications
| Board Dimensions | 100㎜×150㎜ |
|---|---|
| Main Mounted Components | Di1 SoM, various I/F devices |
| Power Supply | 5V from AC adapter or USB-C |
| External Interfaces |
|
| Camera Module | OmniVision OX03C10 module (Hirose DP40) |
| Primary Use | Di1 software development and AI evaluation |
| Operating Environment | 0～45℃ |
Software Development Environment
- ・BSP (RTOS / Linux)
- ・SDK (AI / DSP / Others)
- ・Sample Code
Provided Documentation
- ・User’s Manual
- ・Datasheet
- ・Application Notes
- ・Circuit Diagram
- ・API Reference
- ・Layout Guide
Di1 NPU: Model Zoo (as of June 2026)
Image Classification
- LeNet
- AlexNet
- MobileNet V2
Face Detection
- MTCNNv1 P/R/O NET
- LightFace
Face Recognition
- ArcFace
- Facenet
Pose Estimation
- OpenPose CMU25
- OpenPose CoCO
- MV Recognition
Object Detection
- YOLO V2 COCO 80
- YOLO V3 COCO 80
- YOLO V4 COCO 80
- YOLO V5S6 COCO 80
- YOLO V5N6 TF COCO 80
- YOLO V3 COCO 80 Torch
- YOLO V6S
- YOLO V7 TINY
- YOLO V8S
- YOLO 11n/s/m
- YOLO 12n/s/m
- YOLO 26S
Face Attribute
- Dense Landmarks
- FLD Gaze HP
- 3DDFA
- Facemesh
- L2CS-Net
Audio Classify FFT
- FFT 10sec Parallel
- FFT 2sec Parallel
Image Captioning
- BLIP-image-captioning-base
Semantic Segmentation
- ENET
- Deeplabv3 (resnet50)
- Deeplabv3 (resnet101)
- Deeplabv3 (mobilenet_v3_large)
- Bisenetv1 (cityscapes)
- Bisenetv1 (coco)
Gesture
- Palm Detection
- Hand Landmark
- Gesture Recognition (embededer)
- Gesture Recognition (classifier)
Person Analytics
- CLIP-ViT
More models coming soon
Di1 NPU: ONNX supported operators
Total: 115
| Arithmetic (33): | Add, ArgMax, ArgMin, CumSum, Div, GlobalMaxPool, LogSoftmax, MatMul, Max, MaxPool, MaxRoiPool, Mean, MeanVarianceNormalization, Min, Mod, Mul, NonMaxSuppression, Pow, QLinearMatMul, ReduceL1, ReduceL2, ReduceLogSum, ReduceLogSumExp, ReduceMax, ReduceMean, ReduceMin, ReduceProd, ReduceSum, ReduceSumSquare, Softmax, Sub, Sum, RoiAlign |
|---|---|
| Activation (23): | Atan, Atanh, Celu, Cos, Cosh, Elu, Exp, Expand, Gelu, HardSigmoid, HardSwish, LeakyRelu, Log, Mish, PRelu, Relu, Selu, Sigmoid, Sin, Sinh, Softplus, Softsign, Tanh |
| Tensor Ops (17): | CenterCropPad, Conv, ConvTranspose, Flatten, Gather, GatherElements, Gemm, Pad, QLinearConv, Reshape, Slice, Split, Squeeze, Tile, Transpose, Unsqueeze, ScatterND |
| Normalization & Pooling (7): | AveragePool, BatchNormalization, GlobalAveragePool, GroupNormalization, InstanceNormalization, LRN, LayerNormalization |
| Recurrent (2): | GRU, LSTM |
| Quantization (2): | DequantizeLinear, QuantizeLinear |
| Logical & Cast (16): | And, BitwiseAnd, BitwiseNot, BitwiseOr, BitwiseXor, Cast, CastLike, Equal, Greater, GreaterOrEqual, Identity, Less, LessOrEqual, Or, Where, Xor |
| Other (15): | Abs, Clip, Col2Im, Concat, DepthToSpace, Erf, GridSample, Neg, Reciprocal, Resize, Sign, SpaceToDepth, Sqrt, Tan, TopK |
More operators coming soon
