ITM GLOBAL SCHOOL, GWALIOR
GRADE: X (WOODPECKER/FLICKER)
SUBJECT: Artificial Intelligence(417)
PRACTICAL FILE:
The file should contain a minimum 15 Python programs from Data Science, Natural Language
Processing, and Advance Python
PROJECT FILE:
Project Idea 1: Emotion Detection from Text
Objective: Build a model to classify the emotions expressed in text data (e.g., happy, sad,
angry).
Steps:
1. Data Collection:
○ Gather labeled datasets containing text and corresponding emotions (e.g.,
tweets, movie reviews).
2. Data Preprocessing:
○ Clean and tokenize text data.
○ Use techniques like TF-IDF or word embeddings (e.g., Word2Vec, GloVe).
3. Model Design:
○ Use an LSTM, GRU, or Transformer-based architecture for text classification.
4. Model Training:
○ Train the model and use crossentropy loss for multi-class classification.
5. Evaluation:
○ Test the model and analyze performance using metrics like accuracy and
F1-score.
6. Deployment:
○ Build a web app where users can input text and receive emotion predictions.
Project Idea 2: Air Quality Index Prediction
Objective: Develop a machine learning model to predict air quality indices based on
environmental data.
Steps:
1. Data Collection:
○ Collect historical air quality data (e.g., from government databases or Kaggle).
○ Features may include temperature, humidity, wind speed, and pollutant levels.
2. Data Preprocessing:
○ Handle missing values and outliers.
○ Normalize or standardize features.
3. Model Selection:
○ Choose regression models (e.g., Linear Regression, Random Forest, XGBoost).
4. Model Training:
○ Train the model on historical data and evaluate using metrics like RMSE and R².
5. Evaluation:
○ Analyze model performance with visualizations and comparison against actual air
quality indices.
6. Deployment:
○ Create a dashboard displaying predictions and real-time air quality data.
Project Idea 3: Traffic Sign Recognition
Objective: Build a model to recognize traffic signs in images.
Steps:
1. Data Collection:
○ Use the German Traffic Sign Recognition Benchmark (GTSRB) dataset or create
your own.
2. Data Preprocessing:
○ Resize and normalize images, and split into training, validation, and test sets.
3. Model Design:
○ Use a CNN architecture to classify images of traffic signs.
4. Model Training:
○ Train the model and evaluate with accuracy metrics.
5. Evaluation:
○ Analyze results and visualize the performance on test data.
6. Deployment:
○ Develop an app that allows users to take photos of traffic signs for recognition.
Project Idea 4: Voice Emotion Recognition
Objective: Develop a model to identify emotions from voice recordings.
Steps:
1. Data Collection:
○ Use datasets like the RAVDESS or Emo-DB that contain labeled audio
recordings.
2. Data Preprocessing:
○ Extract audio features using techniques like MFCC (Mel-frequency cepstral
coefficients).
3. Model Design:
○ Use CNNs on spectrogram images or RNNs for sequential data.
4. Model Training:
○ Train the model with appropriate loss functions for classification.
5. Evaluation:
○ Test the model and analyze its performance.
6. Deployment:
○ Create a web app for users to upload audio files and receive emotion
classifications.
Submission Deadline
● Last date for submission is: 30 November 2024
● Submit your project file in both soft copy(pdf) and printed copy.
