Closed Lukeeeeee closed 7 years ago
We use an image size of 256x192, we solved the problem by compressing the image before passing to Python
But the original image is only 64x64. U mean reshape the 64x64 image to 256x192 image?
On Fri, 29 Sep 2017 at 22:31 Ariel Keselman notifications@github.com wrote:
We use an image size of 256x192, we solved the problem by compressing the image before passing to Python
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/imagry/Gym-Torcs-DQN/issues/1#issuecomment-333142582, or mute the thread https://github.com/notifications/unsubscribe-auth/AIvLTF-NeKoY8-EZwJgDz8K7qhVQb-SYks5snP9IgaJpZM4PoUaP .
No, we are using modified torcs, original image is 256x192 as set in the options. It described in readme.
So is there any tutorial about how to use this modified gym torcs? And does it support higher resolution rather than 256x192? Thanks!
@skariel @s271
when run under /Gym-Torcs-DQN/vtorcs-RL-color/
sudo make
got the error below
g++ -I/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/export/include -I/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color -g -O2 -Wall -fPIC -fno-strict-aliasing -O2 -Wno-write-strings -DUSE_RANDR_EXT -DGL_GLEXT_PROTOTYPES -Wall -fPIC -fno-strict-aliasing -O2 -Wno-write-strings -DUSE_RANDR_EXT -DGL_GLEXT_PROTOTYPES -D_SVID_SOURCE -D_BSD_SOURCE -DSHM -DHAVE_CONFIG_H -c TrackData.cpp
g++ -shared -o olethros.so olethros.o driver.o opponent.o spline.o pit.o learn.o strategy.o cardata.o geometry.o Trajectory.o TrackData.o -L/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/export/lib -lalut -lprotobuf -lopencv_core -lopencv_highgui -L/usr/lib -L/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/export/lib -llearning
mkdir /home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/export/drivers/olethros
Exporting olethros.so to /home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/export/drivers/olethros/olethros.so
make[4]: Leaving directory '/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/src/drivers/olethros'
make[4]: Entering directory '/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/src/drivers/scr_server'
make[4]: *** No rule to make target '/usr/lib/gcc/x86_64-linux-gnu/4.8/include/stddef.h', needed by 'scr_server.o'. Stop.
make[4]: Leaving directory '/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/src/drivers/scr_server'
/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/Make-default.mk:680: recipe for target 'subdirs' failed
make[3]: *** [subdirs] Error 1
make[3]: Leaving directory '/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/src/drivers'
/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/Make-default.mk:680: recipe for target 'subdirs' failed
make[2]: *** [subdirs] Error 1
make[2]: Leaving directory '/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/src'
/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color/Make-default.mk:680: recipe for target 'subdirs' failed
make[1]: *** [subdirs] Error 1
make[1]: Leaving directory '/home/dl3-2/donglinsen/Gym-Torcs-DQN/vtorcs-RL-color'
Makefile:30: recipe for target 'restart' failed
make: *** [restart] Error 2
when I use the interface:obs_vision_to_image_rgb_proto(),there are actually some buffer data.However,I just use: cv2.imdecode(image_jpeg, cv2.CV_LOAD_IMAGE_COLOR),the value of image data becomes zeros totally. Have you ever encountered similar problems? Looking forward your respond.Thank you! @skariel @s271
same problem
/home/ye/Desktop/gym_torcs/Gym-Torcs-DQN/vtorcs-RL-color/Make-default.mk:680: recipe for target 'subdirs' failed make[1]: *** [subdirs] Error 1 make[1]: Leaving directory '/home/ye/Desktop/gym_torcs/Gym-Torcs-DQN/vtorcs-RL-color' Makefile:30: recipe for target 'restart' failed make: *** [restart] Error 2
The image given by torcs env only supported 64 64 3, as I know, also mentioned in this issue:
https://github.com/ugo-nama-kun/gym_torcs/issues/2.
And this size, I believe, is too small to train a deep neural network.
So If there any way I can extract a higher resolution image from gym torcs? Or how did you solve this problem? Thanks!