I am very new to perceptilabs and have no idea what specific format of CSV file it needs.
I suspect that the format of my .CSV file is completely wrong, can someone guide me on what to do with my specific dataset.
Thank you
Hello @devittobk, we currently don’t support text data in csv files yet. That might be causing the issue. Looking at the labels for the columns, it looks like you are trying to build an object detection model.
You will have to load the images in one data layer and the labels separately in another layer in the required format. The label file has the shape [Nx Grid_size x Grid_size x (5+num_classes) ]. Assume you have a dataset of 10000 images with 5 different classes and you are dividing each image into 7x7 grid.
Then the shape will be 10000x7x7x10. Here 10 corresponds to [5+number_of_classes]. The 10 dimensions are: bool (presence of object), x_mean, y_mean, x_center, y_center, one_hot_encoding of class_label if there is an object.
Note that there is a bit of info in the PL docs here: https://www.perceptilabs.com/docs/supported_data_files