1.Building a High-Accuracy CNN Model

Challenge: Training a CNN model from scratch while ensuring high accuracy was a significant hurdle.

Solution: We experimented with different architectures, optimized hyperparameters, and used data augmentation techniques to improve model performance.

2.Integrating OpenCV and Audio for Sentiment Analysis

Challenge: Combining OpenCV for facial emotion detection with audio sentiment analysis required synchronizing inputs from both modalities.

Solution: We used time-aligned preprocessing and feature extraction techniques to ensure both inputs contributed meaningfully to sentiment classification.

3.Integrating Cloudflare Workers AI for Sentiment Analysis

Challenge: Deploying AI models on Cloudflare Workers AI while maintaining efficiency and scalability.

Solution: We optimized model inference using Cloudflare’s API, ensuring minimal latency while processing real-time sentiment data.

4.Normalizing Scoring Weights

Challenge: Balancing different scoring metrics for sentiment analysis and depression detection.

Solution: We applied statistical normalization techniques, such as Min-Max scaling, to standardize scores across different models.

5.Prompt Engineering for Chatbot Responses

Challenge: Ensuring the chatbot generates natural and human-like responses.

Solution: We iteratively refined prompts and tested various prompt structures to improve coherence, empathy, and contextual understanding in conversations.