efficientnet_b1

修改说明

针对NPU的修改

简介

EfficientNet模型在ImageNet-1k上以240x240分辨率进行训练。EfficientNet是一种移动友好型纯卷积模型（ConvNet），它提出了一种新的缩放方法，即使用一个简单但高效的复合系数来统一缩放深度、宽度和分辨率的所有维度。

使用方法

pip install -r examples/requirements.txt

import torch
import numpy as np
from ais_bench.infer.interface import InferSession
from ultralytics.models.yolo.detect import DetectionPredictor

# Load a model
om_model = InferSession(0, "./models/yolo11s_bs1.om")

# inference the model
om_conf = OM_Conf()
dp = DetectionPredictor(overrides=om_conf.overrides)
dp.model = om_conf
dp.setup_source("./data")
for dp.batch in dp.dataset:
    paths, im0s, s = dp.batch

    # Preprocess
    im = dp.preprocess(im0s)

    # Inference
    im = np.ascontiguousarray(im).astype(np.float32)  # contiguous
    preds = om_model.infer([im])
    preds_tensor = torch.from_numpy(preds[0])

    # Postprocess
    output = dp.postprocess(preds_tensor, im, im0s)
    print("om result:", output[0].boxes.data)

efficientnet_b1

修改说明

针对NPU的修改

简介

使用方法

pip install -r examples/requirements.txt

import torch
import numpy as np
from ais_bench.infer.interface import InferSession
from ultralytics.models.yolo.detect import DetectionPredictor

# Load a model
om_model = InferSession(0, "./models/yolo11s_bs1.om")

# inference the model
om_conf = OM_Conf()
dp = DetectionPredictor(overrides=om_conf.overrides)
dp.model = om_conf
dp.setup_source("./data")
for dp.batch in dp.dataset:
    paths, im0s, s = dp.batch

    # Preprocess
    im = dp.preprocess(im0s)

    # Inference
    im = np.ascontiguousarray(im).astype(np.float32)  # contiguous
    preds = om_model.infer([im])
    preds_tensor = torch.from_numpy(preds[0])

    # Postprocess
    output = dp.postprocess(preds_tensor, im, im0s)
    print("om result:", output[0].boxes.data)