陳映竹 ( Joanna)
Aged: 26
Gender: female
Education
Master of Science in Public Health, National Defense Medical Center
Advanced fields: Deep Learning, Data visualization, Statistics, Bioinformatics
2019 - 2021
Bachelor of Science in Public Health, Kaohsiung Medical University
Advanced fields: Biostatistics, Epidemiology, Genetics
2015 - 2019
Project Experience
Dec 2022 - Dec 2023
oToBrite Electronics
Obstacle detection of Auto Parking Assist
Object segmentation, PyTorch
- Optimize the model architecture to shorten the computation time by 15%.
- In 380 test videos, the accuracy of parking space status is over 90%.
- The model has reached the mass production specifications.
Dec 2021 - Dec 2022
oToBrite Electronics
Lane Line detection of Lane Departure Warning
Object segmentation, PyTorch
- Improve the model IoU by over 10% and achieved a lane detection rate of 98%, with mass production already implemented for two clients.
- In 270 test videos, the model achieved a lane-switching detection rate higher than 97% and an off-lane distance accuracy exceeding 85%.
Feb 2021 - Jun 2021
Defense Medical Center
Assistive Diagnosis of Posterior Circulation Ischemic Stroke
Object segmentation, Statistics, MxNet
- Propose a novel quantitative scale and enhance the accuracy of prognosis for patients with posterior circulation ischemic stroke.
- Reduce the time required for evaluating patients' prognoses from 30 minutes to less than 5 minutes.
Face detection and identity recognition
Object detection, Verification loss , Apps, MxNet
- Utilize verification loss to quantify similarities among multiple faces for recognizing each face's identity.
- Once users create their own identity profiles, the model can recognize identities without the need for retraining.
Expertise
- Object segmentation
- Object detection
- Statistics
Software
- Python (PyTorch)
- R (MxNet / Shiny-Apps)
