ImageClassification

Image Classification (ailia SDK for Unity)

This folder contains real‑time image classification using ailia SDK. The scene is ImageClassification/ImageClassificationSample.unity, driven by ImageClassification/AiliaImageClassificationSample.cs.

Supported AI Models

• GoogLeNet

  • Purpose: ImageNet‑class image classification.
  • Source: ./AiliaImageClassificationSample.cs
  • Script behavior: Loads googlenet.onnx, sets RGB/CHW/FP32 input format, and queries top‑k classes per frame.

• ResNet‑50

  • Purpose: ImageNet‑class classification.
  • Source: ./AiliaImageClassificationSample.cs
  • Script behavior: Loads resnet50*.onnx according to the resnet50model field (e.g., resnet50.opt.onnx), sets RGB/CHW/INT8 range, and returns ranked logits.

• Inception v3

  • Purpose: ImageNet‑class classification.
  • Source: ./AiliaImageClassificationSample.cs
  • Script behavior: Loads inceptionv3.onnx, sets RGB/CHW/FP32 input, runs per‑frame inference.

Model Files and Paths

• Models download into Application.temporaryCachePath under googlenet, resnet50, or inceptionv3 folders.
• Class labels come from ./AiliaClassifierLabel.cs (English and Japanese names for ImageNet).

Reusing in Your Own Code

• Create a classifier:
  • var clf = new AiliaClassifierModel();
  • Optional: clf.Environment(AILIA_ENVIRONMENT_TYPE_GPU);
  • clf.Settings(format, channelOrder, valueRange);
  • clf.OpenFile(pathToPrototxt, pathToOnnx);
• Run on a Color32[] image:
  • var results = clf.ComputeFromImageB2T(camera, width, height, topK);
  • Map category to labels in AiliaClassifierLabel.

Source Links

• Sample controller: ./AiliaImageClassificationSample.cs
• Labels: ./AiliaClassifierLabel.cs