It is different application using Generative Grasping CNN (GG-CNN) (https://github.com/dougsm/ggcnn).
The GGCNN is developed with an encoder-decoder (kind of autoencoder) models. We implemented the model with different autoencoder models and also a script to evalute the performance of this network using Kinect v2. Also, we make several scripts for training different models conveniently.
We used this dataset for grasping surgical tool,
https://drive.google.com/drive/folders/1KgGC9FO8kf2FxQSSgdHdwDIRfJZLPqTj?usp=share_link
-
Denoising autoencoder
-
Sparse Autoencoder
-
Contractive Autoencoder
-
Variational Autoencoder
-
U-net
-
Ubuntu 16.04
-
Python 3
-
pytorch
For example,
python3 train_ggcnn.py --network ggcnn --dataset cornell --dataset-path /media/aal-ml/Ubuntu_data/olivia/catkin_ws/src/final_olivia --outdir output/ggcnn_model
python3 train_ggcnn_vit.py --network vit_ggcnn --dataset cornell --dataset-path /media/aal-ml/Ubuntu_data/olivia/catkin_ws/src/final_olivia
For your environment,
python3 path-to-the-python --network path-to-the-network --dataset cornell --dataset-path path-to-the-dataset --outdir path-to-the-save folder