n01z3

kv = mx.kvstore.create("local")
dataiter = mx.io.ImageRecordIter(
    path_imgrec=os.path.join(MAIN, fn_bin),
    data_shape=(3, size, size),
    mean_r=128.0,
    mean_g=128.0,
    mean_b=128.0,
    rand_crop=False,
    batch_size=batch_size,
    num_parts=kv.num_workers,

n01z3

import time
import cv2.aruco as A
import numpy as np
import cv2

dictionary = cv2.aruco.getPredefinedDictionary(cv2.aruco.DICT_4X4_50)
board = cv2.aruco.CharucoBoard_create(3,3,.025,.0125,dictionary)

cap = cv2.VideoCapture('../color/sample_3color.avi')

n01z3

# pylint: skip-file
import sys
sys.path.insert(0, "../../python/")

import mxnet as mx
import numpy as np
import logging
import time
import os, sys
import importlib

n01z3

import pandas as pd
import numpy as np
import os
import cv2
import matplotlib.pyplot as plt
import matplotlib.patches as patches
from multiprocessing import Pool
from sklearn.model_selection import StratifiedKFold
from sklearn.model_selection import train_test_split
import random

n01z3

Prevent NVIDIA GPUs' throttling on headless server

• Unlock manual fan & overclock settings
  sudo nvidia-xconfig -a --cool-bits=28 --allow-empty-initial-configuration
• Reboot system
• Create script /usr/local/bin/gpu-fan-control.sh

#!/bin/bash