Published April 11, 2023
| Version v1
Other
Open
Color-Dependent Prediction Stability of Popular CNN Image Classification Architectures
Description
This repository contains the retrained AlexNet, VGG16 and ResNet18 models using additional hue and saturation preprocessing steps, as described in the paper "Color-Dependent Prediction Stability of Popular CNN Image Classification Architectures" by L. Mertens et al., presented at ICANN 2023, Heraklion City, Greece, 26-29 September 2023. It also contains all the analysis CSV files used to generate the results reported in the paper (=ColorAnalysis.zip).
For the corresponding code, see https://gitlab.com/EAVISE/lme/nncolorstabilityanalysis-paper
Acknowledgment: This work was funded by KU Leuven grant IDN/21/010.
Files
ColorAnalysis.zip
Files
(2.4 GB)
| Name | Size | Download all |
|---|---|---|
|
md5:2edcd3d64d4f53babc41e2662487089f
|
244.4 MB | Download |
|
md5:471c865e19d182dc284f444d0c4d3164
|
244.4 MB | Download |
|
md5:e5314f60f69b6c72dd8f1135c38d047d
|
244.4 MB | Download |
|
md5:5ab782ab262947be9ef67773c6c9620f
|
244.4 MB | Download |
|
md5:92c55566f514fef5baa1c8535f5af0bf
|
246.5 MB | Preview Download |
|
md5:4fad60f9a868f2b3f271f291f58ce3cc
|
1.4 kB | Preview Download |
|
md5:3ff7277890c6159efb7acd73bfbbbde4
|
46.8 MB | Download |
|
md5:e13518ad24feb4b37f9aa568a120b62a
|
46.8 MB | Download |
|
md5:428612b4b86e384870ff884167f7ab5d
|
553.4 MB | Download |
|
md5:78b7c23f0368470a4ccf1be094a3da6f
|
553.4 MB | Download |
Additional details
Funding
- KU Leuven
- IDN 21/010
Software
- Repository URL
- https://gitlab.com/EAVISE/lme/nncolorstabilityanalysis-paper
- Programming language
- Python