We combine "cutting-edge AI"
with the execution power to scale globally.
We design technology to be lightweight,
fast, and ready for real-world use.
Why Our Partners Trust Us
Face Reenactment
Able to express a high degree of freedom in head angle and delicate facial expressions through the 3D engine
and deep-learning methods.
3D Motion Capture
3D motion capture enables
real-time and highly accurate full-body tracking
without professional equipment.
Depth Map Upsampling
Utilizes 4K RGB images and a low-resolution depth map to attain
a high-resolution depth map.
Face&Age Recognition
Estimates sex and age through face scan on the Cortex-M7 MCU
(for device personalization usage).
Human Segmentation
Separates a human object
from the background in real-time
from imported images and videos.
Depth Estimation
Approximates relative depth on a single RGB camera
via neural networks, on ARM CPU or mobile GPU.
Single-RGB Bokeh
Produces a natural bokeh effect with a single RGB image.
(With blur intensity control
and focus point function.)
Hair Segmentation
Segregates the hair area in real-time
on ARM CPU, and applies a dyeing effect on mobile GPU.
(OpenGL ES 2.0.)
Real-time Mobile
3D Motion Capture
Highly generalized, achieving high accuracy even in very challenging poses.
Despite being a real-time algorithm running on mobile,
it achieved the world’s best performance on the 3DPW test set.
(March 2024 ~ today)
Procrustes analysis – mean per joint position error (the lower, the better)
Our Algorithm
01
APLA’s real-time algorithm on mobile
achieved higher accuracy than the latest SOTA algorithms using high-spec
PC GPUs.
02
The most widely used accuracy metric
in global 3D human pose estimation papers.
03
Lower PA-MPJPE indicates higher accuracy.
Have a question?
Reach out to our team!
Grow into a global tech leader with APLA’s AI solutions.We're always open to exploring new business opportunities.
이용약관
개인정보 처리방침
에이플라 주식회사 ㅣ APLA Inc.
20, Pangyo-ro 289beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do, Republic of Korea
