keyerror: d_map

[ganya7]

train_d_map = np.asarray(sio.loadmat(train_gts_list[img_idx])['d_map'], dtype=np.float32)

could you please add the code files of data parsing which will create the density map of the files?

[dbsengineering]

you must generated density map and rec your arrays with d_map.
you will find get_density_map_gaussian() function in utils

[chiangc1]

you must generated density map and rec your arrays with d_map.
you will find get_density_map_gaussian() function in utils

@dbsengineering What is that mean?

i make ['d_map'] = utils.get_density_map_gaussian()???

[dbsengineering]

The files already available are only "ground-truth" files. you can open with mat function : np.asarray(sio.loadmat(train_gts_list[img_idx])['image_info'][0, 0][0, 0][0], dtype=np.float32)

But you want use "d_map" in training script. this "d_map" want to be generated by you with gaussian function in utils.py file. Before, to be sure convert image to gray image. Read this article : https://www.researchgate.net/publication/311610470_Single-Image_Crowd_Counting_via_Multi-Column_Convolutional_Neural_Network . All is writing