The pretrained model file seems to be broken.Does any one who has a good one ?

[Genlk]

change a vpn,I have get the unbroken file

[DWJ111]

May I ask if your pre trained model is still available? The model size I downloaded is only 131 KB. Thank you. @Genlk @JiaRenChang

[2c984r83y]

May I ask if your pre trained model is still available? The model size I downloaded is only 131 KB. Thank you. @Genlk @JiaRenChang

The pre-trained model I downloaded is also 131KB, but I am unable to run Test_img.py, same as https://github.com/JiaRenChang/RealtimeStereo/issues/17#issue-1330724237. I notice that the model argument is RTStereoNet, not stackhourglass, you need to chage default argument in Test_img.py

#parser.add_argument('--model', default='stackhourglass', help='select model')
parser.add_argument('--model', default='RTStereoNet',  help='select model')

[flyxuexi]

May I ask if your pre trained model is still available? The model size I downloaded is only 131 KB. Thank you. @Genlk @JiaRenChang

The pre-trained model I downloaded is also 131KB, but I am unable to run Test_img.py, same as #17 (comment). I notice that the model argument is RTStereoNet, not stackhourglass, so just follow run.sh

#!/bin/bash

python main.py --maxdisp 192 \
               --model stackhourglass \
               --datapath dataset/ \
               --epochs 0 \
               --loadmodel ./trained/checkpoint_10.tar \
               --savemodel ./trained/

python finetune.py --maxdisp 192 \
                   --model stackhourglass \
                   --datatype 2015 \
                   --datapath dataset/data_scene_flow_2015/training/ \
                   --epochs 300 \
                   --loadmodel ./trained/checkpoint_10.tar \
                   --savemodel ./trained/

or chage default argument in Test_img.py

#parser.add_argument('--model', default='stackhourglass', help='select model')
parser.add_argument('--model', default='RTStereoNet',  help='select model')

同学你好，能请教一个问题吗。在Test_img.py中，我将stackhourglass改为了RTStereoNet，但下面的内容出现了报错： parser.add_argument('--leftimg', default= './VO04_L.png', help='load model') parser.add_argument('--rightimg', default= './VO04_R.png', help='load model')
这里需要的是什么左右图呀，没有头绪 另外注意到我们的研究方向有些相似，能加个联系方式吗，我是dlut的研究生 感谢

[flyxuexi]

May I ask if your pre trained model is still available? The model size I downloaded is only 131 KB. Thank you. @Genlk @JiaRenChang

The pre-trained model I downloaded is also 131KB, but I am unable to run Test_img.py, same as #17 (comment). I notice that the model argument is RTStereoNet, not stackhourglass, so just follow run.sh

#!/bin/bash

python main.py --maxdisp 192 \
               --model stackhourglass \
               --datapath dataset/ \
               --epochs 0 \
               --loadmodel ./trained/checkpoint_10.tar \
               --savemodel ./trained/

python finetune.py --maxdisp 192 \
                   --model stackhourglass \
                   --datatype 2015 \
                   --datapath dataset/data_scene_flow_2015/training/ \
                   --epochs 300 \
                   --loadmodel ./trained/checkpoint_10.tar \
                   --savemodel ./trained/

or chage default argument in Test_img.py

#parser.add_argument('--model', default='stackhourglass', help='select model')
parser.add_argument('--model', default='RTStereoNet',  help='select model')

我感觉将main.py和finetune.py的model改成stackhourglass不太合适。 因为作者相同的缘故，代码中的stackhourglass和basic是从PSMNet项目移过来的，只有RTStereoNet是本项目新提出的

[2c984r83y]

May I ask if your pre trained model is still available? The model size I downloaded is only 131 KB. Thank you. @Genlk @JiaRenChang

The pre-trained model I downloaded is also 131KB, but I am unable to run Test_img.py, same as #17 (comment). I notice that the model argument is RTStereoNet, not stackhourglass, so just follow run.sh

#!/bin/bash

python main.py --maxdisp 192 \
               --model stackhourglass \
               --datapath dataset/ \
               --epochs 0 \
               --loadmodel ./trained/checkpoint_10.tar \
               --savemodel ./trained/

python finetune.py --maxdisp 192 \
                   --model stackhourglass \
                   --datatype 2015 \
                   --datapath dataset/data_scene_flow_2015/training/ \
                   --epochs 300 \
                   --loadmodel ./trained/checkpoint_10.tar \
                   --savemodel ./trained/

or chage default argument in Test_img.py

#parser.add_argument('--model', default='stackhourglass', help='select model')
parser.add_argument('--model', default='RTStereoNet',  help='select model')

我感觉将main.py和finetune.py的model改成stackhourglass不太合适。 因为作者相同的缘故，代码中的stackhourglass和basic是从PSMNet项目移过来的，只有RTStereoNet是本项目新提出的

没错，你是对的。我之前的回复写错了，我当时没意识到run.sh中的参数是stackhourglass, 这不对，应当改为RTStereoNet. 但是后面把参数改为RTStereoNet这段是没问题的。

#parser.add_argument('--model', default='stackhourglass', help='select model')
parser.add_argument('--model', default='RTStereoNet',  help='select model')

所以我修改了之前的回复，去掉了有关run.sh的内容，保留了将模型修改为RTStereoNet的内容。 左右图就是双目相机采集到的一对图像，你报错可能是因为图像的文件路径不对？你可以换成绝对路径试一下。 这是我的电子邮箱: zhao32450039@gmail.com，请多多指教！

[flyxuexi]

May I ask if your pre trained model is still available? The model size I downloaded is only 131 KB. Thank you. @Genlk @JiaRenChang

The pre-trained model I downloaded is also 131KB, but I am unable to run Test_img.py, same as #17 (comment). I notice that the model argument is RTStereoNet, not stackhourglass, so just follow run.sh

#!/bin/bash

python main.py --maxdisp 192 \
               --model stackhourglass \
               --datapath dataset/ \
               --epochs 0 \
               --loadmodel ./trained/checkpoint_10.tar \
               --savemodel ./trained/

python finetune.py --maxdisp 192 \
                   --model stackhourglass \
                   --datatype 2015 \
                   --datapath dataset/data_scene_flow_2015/training/ \
                   --epochs 300 \
                   --loadmodel ./trained/checkpoint_10.tar \
                   --savemodel ./trained/

or chage default argument in Test_img.py

#parser.add_argument('--model', default='stackhourglass', help='select model')
parser.add_argument('--model', default='RTStereoNet',  help='select model')

我感觉将main.py和finetune.py的model改成stackhourglass不太合适。 因为作者相同的缘故，代码中的stackhourglass和basic是从PSMNet项目移过来的，只有RTStereoNet是本项目新提出的

没错，你是对的。我之前的回复写错了，我当时没意识到run.sh中的参数是stackhourglass, 这不对，应当改为RTStereoNet. 但是后面把参数改为RTStereoNet这段是没问题的。

#parser.add_argument('--model', default='stackhourglass', help='select model')
parser.add_argument('--model', default='RTStereoNet',  help='select model')

所以我修改了之前的回复，去掉了有关run.sh的内容，保留了将模型修改为RTStereoNet的内容。 左右图就是双目相机采集到的一对图像，你报错可能是因为图像的文件路径不对？你可以换成绝对路径试一下。 这是我的电子邮箱: zhao32450039@gmail.com，请多多指教！

谢谢，刚解决就看到回复了,结果确实挺模糊的hh，希望有机会多交流～我的wx发邮箱了