Skip to main content
mooey5775
Projects
mixed_media
Reports
Multi-Sphere, RGB CNN Baseline

Multi-Sphere, RGB CNN Baseline

Should not work as good as the single-sphere case.
Daniel Seita
Created on May 14|Last edited on June 8
Compare with:
https://wandb.ai/mooey5775/mixed_media/reports/BC03-3DoF-1-Sphere-RGB-CNN-Baseline--VmlldzoyMDA0MzU0 for single-sphere case
Update: done, and indeed it's looking pretty bad (room for improvement!). Plots show 0.108 +/- 0.1121 at the end! But why is the eval/MSE_loss looking reasonable? It's not overfitting ... yet the success rate just seems bad? With images, it should clearly see the red vs yellow balls?

Section 1


050100150200Step00.20.40.60.81
BC03_MMMultiSphere_v02_ntrain_0100_cam_rgb_pixel_ee_3DoF_ar_8_hor_100_scaleTarg_2022_05_14_13_40_22_0005
BC03_MMMultiSphere_v02_ntrain_0100_cam_rgb_pixel_ee_3DoF_ar_8_hor_100_scaleTarg_2022_05_14_13_40_22_0004
BC03_MMMultiSphere_v02_ntrain_0100_cam_rgb_pixel_ee_3DoF_ar_8_hor_100_scaleTarg_2022_05_14_13_40_22_0003
BC03_MMMultiSphere_v02_ntrain_0100_cam_rgb_pixel_ee_3DoF_ar_8_hor_100_scaleTarg_2022_05_14_13_40_22_0002
BC03_MMMultiSphere_v02_ntrain_0100_cam_rgb_pixel_ee_3DoF_ar_8_hor_100_scaleTarg_2022_05_14_13_40_22_0001
050100150200Step00.20.40.60.81
Group
Run set
5


Code

https://github.com/mooey5775/softgym_MM/commit/46178f07f71f3ed199564ef89e5fa0d35c40a5a3
https://github.com/Xingyu-Lin/softagent_rpad/commit/8b23b4b60134385200edfe37be175cb23f9cc84f

Created with ❤️ on Weights & Biases.

https://wandb.ai/mooey5775/mixed_media/reports/Multi-Sphere-RGB-CNN-Baseline--VmlldzoyMDA3Njk0
Made with Weights & Biases. Sign up or log in to create reports like this one.