This repository currently provides a link to some of the modelling constraints introduced in--
Kar, K., Kubilius, J., Schmidt, K. M., Issa, E. B., & DiCarlo, J. J. (2019). Evidence that recurrent circuits are critical to the ventral stream's execution of core object recognition behavior. Nature Neuroscience .
You can download the .h5 file from the location below***: click here
*** please note that this file will be soon updated.
Once you download the dataset.h5 file, please check it using matlab (python users can also use this file) in the following way.
>>h5disp('dataset.h5')
HDF5 dataset.h5
Group '/'
Dataset 'i1'
Size: 1320x1
MaxSize: 1320x1
Datatype: H5T_IEEE_F64LE (double)
ChunkSize: []
Filters: none
FillValue: 0.000000
Dataset 'images'
Size: 256x256x3x1320
MaxSize: 256x256x3x1320
Datatype: H5T_IEEE_F64LE (double)
ChunkSize: []
Filters: none
FillValue: 0.000000
Dataset 'obj'
Size: 1320x1
MaxSize: 1320x1
Datatype: H5T_IEEE_F64LE (double)
ChunkSize: []
Filters: none
FillValue: 0.000000
Dataset 'ost'
Size: 1320x1
MaxSize: 1320x1
Datatype: H5T_IEEE_F64LE (double)
ChunkSize: []
Filters: none
FillValue: 0.000000
The objects are numbered 1-10. Here's the lookup table
1. 'bear'
2. 'elephant'
3. 'person'
4. 'car'
5. 'dog'
6. 'apple'
7. 'chair'
8. 'plane'
9. 'bird'
10. 'zebra'