This project focuses on image classification using the Devanagari Handwritten Character Dataset (DHCD) for Nepali characters. Leveraging a pretrained ResNet model, the goal is to achieve accurate character recognition.
- Utilizes deep learning techniques for image classification.
- Training is performed on the DHCD (Nepali) dataset.
- Fine-tunes the ResNet model to improve its recognition capabilities.
- The model is designed to recognize handwritten Nepali characters.
📖 Dataset used in this project can be found on DHCD
📜 ResNet101 undergoes transfer learning for the image classification process.
- Fine-tuning of ResNet101 is performed for the image classification task, following are the results after 5 epochs:
- Test Accuracy Score: 0.9922
- Train Accuracy Score: 0.8265
Most Confused Character Pairs:
- character_14_dhaa misclassified as character_18_da: 12 times
- character_23_ba misclassified as character_29_waw: 7 times
- character_29_waw misclassified as character_16_tabala: 6 times
- character_17_tha misclassified as character_26_yaw: 4 times
- character_4_gha misclassified as character_21_pa: 3 times
- digit_6 misclassified as digit_3: 3 times
- character_11_taamatar misclassified as character_16_tabala: 2 times
- character_12_thaa misclassified as character_11_taamatar: 2 times
- character_17_tha misclassified as character_4_gha: 2 times
- character_18_da misclassified as character_14_dhaa: 2 times
- character_19_dha misclassified as character_17_tha: 2 times
- character_24_bha misclassified as character_10_yna: 2 times
- character_2_kha misclassified as character_32_patalosaw: 2 times
- character_33_ha misclassified as character_13_daa: 2 times
- character_35_tra misclassified as character_23_ba: 2 times
- character_5_kna misclassified as character_13_daa: 2 times
- digit_3 misclassified as digit_6: 2 times
- digit_5 misclassified as character_27_ra: 2 times
- character_10_yna misclassified as character_8_ja: 1 times
- character_11_taamatar misclassified as character_13_daa: 1 times
