Skip to content

Instantly share code, notes, and snippets.

View ClementPinard's full-sized avatar

Clément Pinard ClementPinard

View GitHub Profile
import torch, kornia, h5py, imageio
import matplotlib.pyplot as plt
import numpy as np
# This is a failure mode, with different aspect ratios
fname1 = '47698078_3766965066'
fname2 = '18698491_4586522698'
# This works
# fname1 = '271147142_778c4e7999_o'
from __future__ import division
from __future__ import print_function
import argparse
import math
import time
from tqdm import trange
import torch
from correlation_package.correlation import Correlation
@ClementPinard
ClementPinard / bench.py
Created June 12, 2018 11:25 — forked from zou3519/bench.py
[pytorch] GridSampler CUDNN vs THCUNN performance comparision script
import time
import torch
import torch.backends.cudnn as cudnn
import torch.nn.functional as F
from torch.autograd import Variable
def benchmark_shape(N, C, IH, IW, H, W, nrand, nrep):
"""
@ClementPinard
ClementPinard / Flow Augmentation.md
Created January 17, 2018 16:02
Flow Augmentation

given $\theta_1$ and $\theta_2$ as affine transformations for $I_1$ and $I_2$ $$ \theta_1 = \left(\begin{array}{ccc}r_{0,1}&r_{1,1}&t_{0,1}\r_{2,1}&r_{3,1}&t_{1,1}\end{array} \right) = \left( R_1 | T_1 \right) $$ $$ \theta_2 = \left(\begin{array}{ccc}r_{0,2}&r_{1,2}&t_{0,2}\r_{2,2}&r_{3,2}&t_{1,2}\end{array} \right) = \left( R_2 | T_2 \right) $$

$$

@ClementPinard
ClementPinard / Still_Box_README.md
Last active September 19, 2017 10:58
Still Box README

Still Box

Torrent description

In addition to this README, this torrent contains 4 datasets:

Name Image size (px) Scene number Size compressed (B) Total size (B)
64.tar.xz 64x64 80K 9.8G 19G
128.tar.xz 128x128 20K 7.1G 12G