部署训练出来的模型后，满屏都是识别到了物体，画面卡顿厉害

import cmath
import gc
import time

import image
import lcd
import sensor
from maix import KPU

lcd.init()  # Init lcd display
lcd.clear(lcd.RED)  # Clear lcd screen.

# sensor.reset(dual_buff=True)      # improve fps
sensor.reset()  # Reset and initialize the sensor.
sensor.set_pixformat(sensor.RGB565)  # Set pixel format to RGB565 (or GRAYSCALE)
sensor.set_framesize(sensor.QVGA)  # Set frame size to QVGA (320x240)
sensor.set_vflip(True)  # 翻转摄像头
sensor.set_hmirror(True)  # 镜像摄像头
sensor.skip_frames(time=1000)  # Wait for settings take effect.
clock = time.clock()  # Create a clock object to track the FPS.

print("ready load model")

labels = ["apple"]  # 类名称，按照label.txt顺序填写
anchor = (2.61, 2.95, 3.99, 2.27, 4.75, 3.43, 4.32, 4.78, 7.32, 5.68)  # anchors,使用anchor.txt中第二行的值

kpu = KPU()
# 从sd或flash加载模型
# kpu.load_kmodel('/sd/det.kmodel')
kpu.load_kmodel(0x300000, 584744)
kpu.init_yolo2(
    anchor,
    anchor_num=(int)(len(anchor) / 2),
    img_w=320,
    img_h=240,
    net_w=320,
    net_h=240,
    layer_w=10,
    layer_h=8,
    threshold=0.6,
    nms_value=0.3,
    classes=len(labels),
)

while True:
    gc.collect()

    clock.tick()
    img = sensor.snapshot()

    kpu.run_with_output(img)
    dect = kpu.regionlayer_yolo2()

    fps = clock.fps()

    if len(dect) > 0:
        for l in dect:
            a = img.draw_rectangle(l[0], l[1], l[2], l[3], color=(0, 255, 0))

            info = "%s %.3f" % (labels[l[4]], l[5])
            a = img.draw_string(l[0], l[1], info, color=(255, 0, 0), scale=2.0)
            print(info)
            del info

    a = img.draw_string(0, 0, "%2.1ffps" % (fps), color=(0, 60, 255), scale=2.0)
    lcd.display(img)

试试看这个脚本，而且看你的训练结果，测试数据都没有几个正确框出来的。

你好，识别使用的nms和threasold还有anchor需要按照模型训练结果中的配置。