Face detection or face recognition by using deep learning algorithms
Budget: $30 – $250 USD
1. First, select the topic that interests you. Task should be challenging enough (e.g., Image segmentation, object detection, face recognition) to qualify for the final project. Simple classification/regression task is not allowed.
2. Do a search to find quality papers. You can search on google scholars,, IEEEXplore (https://unh-proxy01.newhaven.edu/login?url=https://www.ieee.org/ieeexplore). Top conferences include NeurIPS, CVPR, ECCV, ICCV, ICML.
3. Once you find candidate papers, do a literature survey (read the abstract, intro, and results) to pick the one that you can work, Find online PyTorch code.
4. Read the dataset part carefully. Look into their code and dataset to find how the data needed to be formatted.
5. You need to build your own dataset. Select the source that you will be using and provide few samples as examples in the proposal. Mention if you will be using any tools for annotation. The source should be something that possibly have not been used before for building dataset. For example, pictures taken on campus that can be easily recognized. You can also use stable diffusion to generate images from text. If you use this, you must provide all the prompts that you used to generate those images and you should provide snapshots. The stable diffusion model can be downloaded from https://github.com/CompVis/stable-diffusionLinks to an external site.. There are other memory and computationally efficient version. Make sure that you have sufficient hardware capacity to run this model.
6. Your report should include title (which should be the same as the title of the paper), name of the team member, a short description of the the work/method, description of the dataset and format, data source, annotation tools. 2/3 samples (that you will create as part of the proposal). Also, attach the paper and provide links to online code, data source.
2. Do a search to find quality papers. You can search on google scholars,, IEEEXplore (https://unh-proxy01.newhaven.edu/login?url=https://www.ieee.org/ieeexplore). Top conferences include NeurIPS, CVPR, ECCV, ICCV, ICML.
3. Once you find candidate papers, do a literature survey (read the abstract, intro, and results) to pick the one that you can work, Find online PyTorch code.
4. Read the dataset part carefully. Look into their code and dataset to find how the data needed to be formatted.
5. You need to build your own dataset. Select the source that you will be using and provide few samples as examples in the proposal. Mention if you will be using any tools for annotation. The source should be something that possibly have not been used before for building dataset. For example, pictures taken on campus that can be easily recognized. You can also use stable diffusion to generate images from text. If you use this, you must provide all the prompts that you used to generate those images and you should provide snapshots. The stable diffusion model can be downloaded from https://github.com/CompVis/stable-diffusionLinks to an external site.. There are other memory and computationally efficient version. Make sure that you have sufficient hardware capacity to run this model.
6. Your report should include title (which should be the same as the title of the paper), name of the team member, a short description of the the work/method, description of the dataset and format, data source, annotation tools. 2/3 samples (that you will create as part of the proposal). Also, attach the paper and provide links to online code, data source.