litert-community/MobileNet-v2

MobileNet V2

MobileNet V2 model pre-trained on ImageNet-1k at resolution 224x224. It was introduced in MobileNetV2: Inverted Residuals and Linear Bottlenecks by Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, Liang-Chieh Chen.

Model description

The model was converted from a checkpoint from PyTorch Vision.

The original model has:
acc@1 (on ImageNet-1K): 71.878%
acc@5 (on ImageNet-1K): 90.286%
num_params: 3,504,872

The license information of the original model was missing.

Available model files

File	Description
mobilenet_v2.tflite	Full precision LiteRT/TFLite model.
mobilenet_v2_dynamic_wi8_afp32.tflite	Dynamic weight-only INT8 model with FP32 activations.
mobilenet_v2_Google_Tensor_G5_apply_plugin.tflite	AOT-compiled artifact for the Google Tensor G5 target.
mobilenet_v2_int8_channelwise.tflite	Static INT8 model with channelwise INT8 weights and asymmetric INT8 activations.

Quantization

mobilenet_v2_int8_channelwise.tflite was produced with the STATIC_WI8_AI8 quantization recipe. Weights are signed INT8 and use symmetric channelwise quantization for weight tensors. Activations are signed INT8 with asymmetric quantization parameters.

The INT8 channelwise artifact keeps standard LiteRT/TFLite model structure before AOT compilation. Local LiteRT compiler checks fully delegated this artifact on tested compatible NPU backends. Enablement for other NPU backends is still under validation.

Use

#!/usr/bin/env python3
import argparse, json
import numpy as np
from PIL import Image
from huggingface_hub import hf_hub_download
from ai_edge_litert.compiled_model import CompiledModel

def preprocess(img: Image.Image) -> np.ndarray:
    img = img.convert("RGB")
    w, h = img.size
    s = 256
    if w < h:
        img = img.resize((s, int(round(h * s / w))), Image.BILINEAR)
    else:
        img = img.resize((int(round(w * s / h)), s), Image.BILINEAR)
    left = (img.size[0] - 224) // 2
    top = (img.size[1] - 224) // 2
    img = img.crop((left, top, left + 224, top + 224))

    x = np.asarray(img, dtype=np.float32) / 255.0
    x = (x - np.array([0.485, 0.456, 0.406], dtype=np.float32)) / np.array(
        [0.229, 0.224, 0.225], dtype=np.float32
    )
    return np.transpose(x, (2, 0, 1))

def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("--image", required=True)
    args = ap.parse_args()

    model_path = hf_hub_download("litert-community/MobileNet-v2", "mobilenet_v2.tflite")
    labels_path = hf_hub_download(
        "huggingface/label-files", "imagenet-1k-id2label.json", repo_type="dataset"
    )
    with open(labels_path, "r", encoding="utf-8") as f:
        id2label = {int(k): v for k, v in json.load(f).items()}

    img = Image.open(args.image)
    x = preprocess(img)

    model = CompiledModel.from_file(model_path)
    inp = model.create_input_buffers(0)
    out = model.create_output_buffers(0)

    inp[0].write(x)
    model.run_by_index(0, inp, out)

    req = model.get_output_buffer_requirements(0, 0)
    y = out[0].read(req["buffer_size"] // np.dtype(np.float32).itemsize, np.float32)

    pred = int(np.argmax(y))
    label = id2label.get(pred, f"class_{pred}")

    print(f"Top-1 class index: {pred}")
    print(f"Top-1 label: {label}")
if __name__ == "__main__":
    main()

BibTeX entry and citation info

@inproceedings{mobilenetv22018,
  title={MobileNetV2: Inverted Residuals and Linear Bottlenecks},
  author={Mark Sandler and Andrew Howard and Menglong Zhu and Andrey Zhmoginov and Liang-Chieh Chen},
  booktitle={CVPR},
  year={2018}
}