cGAN Model Review & Fine-Tuning for Image Generation Task
Budget: ₹750 – ₹1,250 INR
I have a Conditional GAN (cGAN) implemented in PyTorch that generates images from one-hot encoded Arabic text inputs. The model is operational, but I’m seeking a deep learning professional with expertise in GANs, to review the model’s architecture, training setup, and results, and provide fine-tuning suggestions for improved image quality and performance.
Project Details:
Dataset: 76,000 Arabic word images, 200x200 in size, with corresponding one-hot encoded text vectors.
Model: The model consists of a Generator and a Discriminator for the Conditional GAN (cGAN) setup. I have integrated noise vectors with the one-hot encoded text and trained the model with a Binary Cross Entropy Loss function.
Current Status: The training pipeline is set up with checkpoints saved, and images generated at regular intervals. The code runs successfully, but I need to improve the quality of generated images and optimize the training process.
Key Tasks:
- Code Review: Review the Generator and Discriminator model architectures, loss function, optimizers, and schedulers.
Fine-Tuning: Provide suggestions for hyperparameter tuning, architecture adjustments, or any techniques that may enhance the performance and produce realistic images.
Skills and Experience Required:
- Proficiency in PyTorch and GANs
- Strong background in image processing
- Experience in fine-tuning machine learning models
- Ability to generate realistic images from a GAN
I’m looking to get started today with quick feedback and suggestions.
Project Details:
Dataset: 76,000 Arabic word images, 200x200 in size, with corresponding one-hot encoded text vectors.
Model: The model consists of a Generator and a Discriminator for the Conditional GAN (cGAN) setup. I have integrated noise vectors with the one-hot encoded text and trained the model with a Binary Cross Entropy Loss function.
Current Status: The training pipeline is set up with checkpoints saved, and images generated at regular intervals. The code runs successfully, but I need to improve the quality of generated images and optimize the training process.
Key Tasks:
- Code Review: Review the Generator and Discriminator model architectures, loss function, optimizers, and schedulers.
Fine-Tuning: Provide suggestions for hyperparameter tuning, architecture adjustments, or any techniques that may enhance the performance and produce realistic images.
Skills and Experience Required:
- Proficiency in PyTorch and GANs
- Strong background in image processing
- Experience in fine-tuning machine learning models
- Ability to generate realistic images from a GAN
I’m looking to get started today with quick feedback and suggestions.