Hi, we think the task is reasonable. Since there are publicly available solutions (even on Kaggle) you should add something you did yourself. Some suggestions: - Augment the training dataset (learn what this is and how to do it). - Utilize a pretained face recognition/detection or object detection model (transfer learning). - Find another dataset of facial keypoints or face detection and check if you can merge or otherwise use it. Do a short literature research and check if you can find something which is easy to implement. State sources in your presentation. If you got inspired by a blog post, article etc. reference this. You can use additional (Python) packages, if you want. I would estimate that the training takes hours, so start early. Kind regards, Jannik Schürg On 05.07.19 13:52, Elizabeth Baker wrote:

Hi,

For our final project we wanted to use the 'Facial Keypoints Detection' dataset on kaggle:

https://www.kaggle.com/c/facial-keypoints-detection/overview

Here we would be trying to use the data to predict where features lie on a persons face. We wanted to first check with you to make sure that this task is reasonable, since it is not one of the tasks on the sheet.

The data files are described as follows:* * *Data files*

* *training.csv:* list of training 7049 images. Each row contains the (x,y) coordinates for 15 keypoints, and image data as row-ordered list of pixels. * *test.csv:* list of 1783 test images. Each row contains ImageId and image data as row-ordered list of pixels * *submissionFileFormat.csv:* list of 27124 keypoints to predict. Each row contains a RowId, ImageId, FeatureName, Location. FeatureName are "left_eye_center_x," "right_eyebrow_outer_end_y," etc. Location is what you need to predict.

Thank you for your input - we hope this will be an appropriate project!

Best wishes, Elizabeth Baker and Jan Wuzyk

<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail> Virus-free. www.avast.com <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>

-- -- Jannik Schürg Address: INS 1.046, Endenicher Allee 19b, 53115 Bonn Tel.: +49 228 73-69753