Fine-Tune AI Model for Symbol Recognition in Architectural Drawings -- 2
Budget / Salary$250–500
TypeFreelance project
LocationRemote
Posted1 hour ago
I am seeking an experienced AI programmer to help complete and improve an existing AI model designed to recognize and quantify symbols in architectural PDF drawings, including accurately separating individual buildings during the output process.
The current model struggles to accurately identify and count certain symbols, resulting in missed symbols, duplicate counts, and incorrect quantities. The goal is to assess its current performance, identify the underlying issues, and improve its recognition, counting, and building separation capabilities.
The model must achieve a 95–98% accuracy rate in symbol recognition, quantity counting, and building separation.
Required Deliverables:
1. Current Accuracy Assessment: Evaluate the existing model's accuracy, identify recurring errors, and establish a performance baseline.
2. Troubleshooting and Improvements: Identify and resolve issues affecting symbol recognition, quantity counting, and building separation.
3. Symbol Recognition: Improve the model's ability to identify and distinguish different symbols, including similar-looking ones.
4. Accurate Counting: Ensure symbols are counted correctly, avoiding missed or duplicate counts.
5. Building Separation: Ensure the model can accurately identify and separate individual buildings in architectural drawings, with quantities and results correctly assigned to each building.
6. Testing and Validation: Test the improved model using sample architectural drawings and demonstrate the achievement of the 95–98% accuracy target.
7. Final Delivery: Provide the updated functional model, relevant source code, and documentation of the improvements and test results.
Required Skills and Experience:
The freelancer must have extensive hands-on coding experience with Python, OpenAI technologies, and Laravel. Strong expertise in AI development, machine learning, model training and fine-tuning, computer vision, image processing, and object detection is essential. Experience working with architectural or technical drawings is highly preferred.
The current model struggles to accurately identify and count certain symbols, resulting in missed symbols, duplicate counts, and incorrect quantities. The goal is to assess its current performance, identify the underlying issues, and improve its recognition, counting, and building separation capabilities.
The model must achieve a 95–98% accuracy rate in symbol recognition, quantity counting, and building separation.
Required Deliverables:
1. Current Accuracy Assessment: Evaluate the existing model's accuracy, identify recurring errors, and establish a performance baseline.
2. Troubleshooting and Improvements: Identify and resolve issues affecting symbol recognition, quantity counting, and building separation.
3. Symbol Recognition: Improve the model's ability to identify and distinguish different symbols, including similar-looking ones.
4. Accurate Counting: Ensure symbols are counted correctly, avoiding missed or duplicate counts.
5. Building Separation: Ensure the model can accurately identify and separate individual buildings in architectural drawings, with quantities and results correctly assigned to each building.
6. Testing and Validation: Test the improved model using sample architectural drawings and demonstrate the achievement of the 95–98% accuracy target.
7. Final Delivery: Provide the updated functional model, relevant source code, and documentation of the improvements and test results.
Required Skills and Experience:
The freelancer must have extensive hands-on coding experience with Python, OpenAI technologies, and Laravel. Strong expertise in AI development, machine learning, model training and fine-tuning, computer vision, image processing, and object detection is essential. Experience working with architectural or technical drawings is highly preferred.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.