polynomialqian / rd3d Goto Github PK

The pkl file generate by myself which use the code from the internet got something different compared with yours. Could you please send me the code which you used to generate the pkl file. Thank very much!

IN[1]: runfile('C:/Users/Asus/Desktop/RD3D-master/train.py', wdir='C:/Users/Asus/Desktop/RD3D-master')

Namespace(batchsize=10, clip=0.5, data_dir='', epochs=50, hflip=False, lr=0.000125, lr_decay_epochs=[120, 160, 200], lr_decay_rate=0.1, lr_decay_steps=20, lr_scheduler='cosine', momentum=0.9, optim='adamW', output_dir='./output\1630156223', trainsize=352, vflip=False, warmup_epoch=-1, warmup_multiplier=100, weight_decay=0.0001)
[08/28 17:40:23 rd3d]: Full config saved to ./output\1630156223\config.json
========>num_gpus:1==========

Traceback (most recent call last):

File "C:\Users\Asus\Desktop\RD3D-master\train.py", line 169, in
ckpt_path = main(opt)

File "C:\Users\Asus\Desktop\RD3D-master\train.py", line 77, in main
train_loader = build_loader(opt)

File "C:\Users\Asus\Desktop\RD3D-master\train.py", line 60, in build_loader
train_loader = get_loader(image_root, gt_root, depth_root, batchsize=opt.batchsize * num_gpus,

File "C:\Users\Asus\Desktop\RD3D-master\data.py", line 127, in get_loader
dataset = SalObjDataset(image_root, gt_root, depth_root, trainsize, hflip=hflip, vflip=vflip)

File "C:\Users\Asus\Desktop\RD3D-master\data.py", line 23, in init
self.filter_files()

File "C:\Users\Asus\Desktop\RD3D-master\data.py", line 83, in filter_files
assert len(self.images) == len(self.gts)

AssertionError

File "C:\Users\Asus\anaconda3\lib\site-packages\PIL\Image.py", line 2891, in open
fp = builtins.open(filename, "rb")

FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\Asus\Desktop\RD3D-master\NJU2K\jpg000003_left.jpg'

I have 5000 pairs of images from stereo camera kit? I am wondering is it possible to use my data and you model to train a regression prediction model? All my depth images are from stereo camera kit and generated from PSMnet.

Best,
Zhongjie

Cannot find the paper everywhere, including github, google scholar, and Baidu scholar. Could you please share me the address?

作者你好，我现在有一个项目，需要测试你的模型，请问你能发布训练好的模型吗？我的邮箱是[email protected]，万分感谢！！！

It's invalid now

您好，根据论文设定，增广是随机水平翻转和多尺度{256，352，416}，但是复现出来的结果和给出的结果相差比较大；使用github代码，增广还包括了裁剪、旋转和椒盐噪声，多尺度是{256，352，448=round(352*1.25/32)*32}，但是复现的结果部分数据集仍相差较大，当然DL不可能完全复现，同时椒盐噪声提高鲁棒性同时对模型性能影响不同。请问您训练时候采用的增广方式有哪些呢

Traceback (most recent call last):
File "", line 1, in
File "C:\Users\Asus\anaconda3\lib\multiprocessing\spawn.py", line 116, in spawn_main
exitcode = _main(fd, parent_sentinel)
File "C:\Users\Asus\anaconda3\lib\multiprocessing\spawn.py", line 126, in _main
self = reduction.pickle.load(from_parent)
EOFError: Ran out of input
Traceback (most recent call last):
File "", line 1, in
File "C:\Users\Asus\anaconda3\lib\multiprocessing\spawn.py", line 116, in spawn_main
exitcode = _main(fd, parent_sentinel)
File "C:\Users\Asus\anaconda3\lib\multiprocessing\spawn.py", line 126, in _main
self = reduction.pickle.load(from_parent)
MemoryError

Hi, 我在训练的第一阶段遇到了问题，你可以帮我解决吗？

作者您好，使用您提供的预训练模型在自己网上找的同名数据集上评估结果比较差，请问您可以提供一下您使用的数据集么？
谢谢！邮箱[email protected]

作者你好，我这几天使用您提供的代码重新训练了好几次，我使用的是单张TitanXp，batch_size设为8（10的话显卡放不下），epochs=100，其他参数基本都使用您代码中的默认参数，最终训练loss都收敛到0.03左右，但测试效果都很不理想。所以我想请教一下您训练时的具体训练参数和最终loss收敛值是多少，最好能提供一下训练好的模型文件？
万分感谢！

Hi, thanks for sharing.
Can you provide the RD3D results of ReDWeb-S?