Optical Tactile (TacTip) Dataset for texture classification

This dataset provides a comprehensive collection of temporal images captured by an optical tactile sensor (TactiP) mounted on a converted CNC machine, aimed at facilitating texture classification research in robotics and manufacturing applications. The dataset comprises high-resolution temporal image sequences acquired while the sensor was dragged across a diverse range of textures. Each sequence is meticulously labeled with corresponding texture categories, encompassing various materials and surface patterns to ensure a broad and representative coverage of textural variations. The dataset is designed to support the development and evaluation of advanced texture classification algorithms, offering insights into the dynamic interactions between tactile sensors and textured surfaces. By providing both spatial and temporal dimensions of texture data, this dataset aims to enhance the understanding of texture perception and improve the performance of tactile sensing systems in practical applications.About this fileThis is a dataset made up of optical tactile sensors being dragged across a surface at varying pressures. The dataset contains different readings across 20 frames, all converted to grey-scale. We have two parts of this dataset "X_data_15" and "X_data_gel_15". The first one is a sensor that uses clear silicone, and the second makes use of a clear gel. The gel is far more sensitive but prone to damage. The clear silicone is less sensitive but very resilient to pressure and damage.The labels are as follows and the index corresponds to the number in the y data:**Labels:**+-------+---------------+----------+| Index | Label | Friction Value |+-------+---------------+----------+| 0 | Carpet | - || 1 | LacedMatt | 0.124083 || 2 | wool | 0.198984 || 3 | Cork | 0.344905 || 4 | Felt | 0.116275 || 5 | LongCarpet | 0.128852 || 6 | cotton | 0.129354 || 7 | Plastic | 0.101801 || 8 | Flat | 0.343766 || 9 | Ffoam | 0.396798 || 10 | Gfoam | 0.191812 || 11 | bubble | 0.217116 || 12 | Efoam | 0.073127 || 13 | jeans | - || 14 | Leather | 0.255904 |+-------+---------------+----------+Each X item is an image with a set number of frames recording in real time at approximately 100 FPS. The y labels is a number corresponding to the index of the labels shown above. FOr example, 0 would be a carpet.X shape: (Number, frame, h, w)y shape: (Number,)All files are in compressed numpy format. Python users can load in the dataset using the code provided in the ReadMe.