search

SrijanShovit / HealthLearning

A repo comprising of various Machine Learning and Deep Learning projects in healthcare domain.
39 stars 53 forks source link

Acne Prediction | Input pipeline and Dataset preprocessing #49

Closed Saksh8 closed 6 months ago

Saksh8 commented 6 months ago

I wish to add a deep learning model that could predict predict acne based on the image analysis

SrijanShovit commented 6 months ago

Ok explain the plan like dataset and further approaches. I shall then assign to you.

Saksh8 commented 6 months ago

Dataset is on Kaggle containing 927 image files labeled using Roboflow in YOLOv8 format. This dataset was divided into train, validation, and test sets, and data replication processes were applied.

To implement my model, I decided to base it on the YOLOv8 architecture, drawing inspiration from the Keras example. This architecture is well-suited for object detection tasks like identifying acne in images.

One of the key components of my approach was leveraging TensorFlow's tf.data API for building an efficient input pipeline. This allowed me to preprocess the dataset and create batches of training data seamlessly, optimizing the training process.

Since the dataset was labeled in YOLOv8 format, I crafted a parser function to read and interpret the bounding box annotations and class labels from the accompanying text files.

With my input pipeline ready and the YOLOv8 model architecture defined, I embarked on the training phase. I monitored the training process closely, using the validation data to gauge the model's performance and make adjustments as needed.

After training, I will evaluate the model's performance on the test data to assess its accuracy and generalization capabilities.

SrijanShovit commented 6 months ago

@Saksh8 Perfect, let's proceed in steps. Each step has its separate issue and thus separate points.

Input pipeline and Dataset preprocessing -- noise removal, augmentation, etc. -> Level 2 (continue this issue only) Next take up different models like custom CNN, Dilated CNN, Pre-trained models, Graph CNN, etc. We will have separate issue for each type of model.

SrijanShovit commented 6 months ago

Make sure to add proper readme in separate folder for Acne.

Saksh8 commented 6 months ago

Sure.Thank You

Saksh8 commented 6 months ago

I have resolved this issue also .

SrijanShovit commented 6 months ago

I have resolved this issue also .

ok will have a look today.

github-actions[bot] commented 6 months ago

This issue has been automatically closed because it has been inactive for more than 7 days. If you believe this is still relevant, feel free to reopen it or create a new one. Thank you!