Fig 1.
Block diagram of proposed method.
Fig 2.
Normal and Ulcer foot images in merged DFUC 2021.
Fig 3.
Data Distribution of DFUC2021, DFUC2020 and Kaggle DFU Datasets.
Table 1.
Dataset distribution.
Table 2.
Preprocessing and data augmentation techniques with parameters.
Table 3.
Model parameters.
Fig 4.
Architecture of proposed DFU_DIALNet approach.
Table 4.
Result evaluation of all models in DFUC2021 dataset.
Table 5.
Classification report of different models in DFUC2021 dataset.
Fig 5.
Train vs Test accuracy plot and Confusion Matrix of DFU_DIALNet in DFUC2021.
Table 6.
Result Evaluation of DFU_DIALNet model on KDFU and DFUC2020 datasets.
Fig 6.
Performance evaluation of DFU_DIALNet: Train vs Test Accuracy plot in DFUC2020 and KDFU Dataset.
Fig 7.
Confusion Matrix of DFU_DIALNet in KDFU & DFUC2020 dataset.
Fig 8.
Receiver Operating Characteristic curve of DFU_DIALNet in DFUC2021, KDFU & DFUC2020 datasets.
Fig 9.
Result Comparison of DFU_DIALNet in DFUC2021, KDFU & DFUC2020 datasets.
Table 7.
Comparison of performance across various research studies and proposed DFU_DIALNet approach on the DFUC2021, KDFU & DFUC2020 datasets.
Fig 10.
GradCam Explanation with DFU_DIALNet.
Fig 11.
LIME Explanation with DFU_DIALNet.
Fig 12.
GradCam Visualization of high performing models in Ulcer Foot images.
Fig 13.
GradCam Visualization of high performing models in Normal Foot images.
Fig 14.
Developed WebApp to detect DFU images.