Manufacturing processes have limitations that cause deviations and deformities, thus require inspections. Current surface finishes inspections rely on custom processes with expensive tools or human labor.
We aim to fix that with inspeksi, the smart defect detector. Our plug and play solution offers a wide range of automated features to make inspection of newly manufactured parts fast, efficient, and effective. Simply insert your part into the enclosure and watch how the automated system scans your part and detects defects using a sophisticated AI model that only increases in accuracy as you continue to use the product.
What makes us great
We've made inspeksi easy to use by letting it do all the work for you! Our sytem runs at the press of a button giving you all the details on the defects on your product.
Our defect detection algorithm is highly accurate, and even becomes more effective as the product learns about defects while you use it.
Our device is customizable to your needs. You can save the positions that you want to scan on your part using our record and play back feature.
Find out if a part is defective in seconds. Our system scans several parts a minute, and gives you the results you need almost instantly.
Specify what you want inspeksi to consider a defective part, and what defects are acceptable using our interactive user interface.
All you need to get started is two 120v outlets, and a part that you want to inspect, simple as that!
Meet the guys behind inspeksi.
Robot Design Lead
Product Design and Integration Lead
The latest news about inspeksi.
Mar. 13 '19 Herman Grewal 5 min
Finishing off our Symposium Poster
Here's our poster for the symposium! We've gone for a bright and exciting look to attract eyes on Friday.
Mar. 12 '19 Lalit Lal 2 min
Running Everything Together
Everything has been fully integrated and now it's time for some final tests.
Mar. 10 '19 Rey Wiyatno 2 min
Testing our Defect Detection Algorithm on real parts
We're using parts donated by ecobee in order to test our system's effectiveness using a dataset containing completely arbitrary parts.
Feb. 24 '19 Oluwatoni Ogunmade 4 min
Recording arm positions and playback
Utilizing the back-driveability of the arms to allow positions to be recorded and played back.
Feb. 17 '19 Oluwatoni Ogunmade 5 min
Robotic Arm Redesigns
Changes made to the initial design of the robotic arm after manufacturing and assembly to reduce the arm's weight and improve the strength.
Feb. 7 '19 Lalit Lal 3 min
Completing Electrical Assembly of Robotic Arm
The team is using various electronics that operate at different voltage levels. Since various parts communicate with each other, some interfacing and electronic design is required.
Feb. 5 '19 Lalit Lal 3 min
Establishing Communication with Dynamixel Smart Serial Servos
Communicating with the smart servos is a crucial stage in development to initiate the controls software development for moving the robotic arm in a confined space.
Feb. 3 '19 Rey Wiyatno 5 min
From Object Detection to Instance Segmentation
The algorithm used for the initial demo in the final design project was the object detection algorithm called the Single-shot Multibox Detector (SSD). The algorithm is now changed to instance segmentation model called the Mask R-CNN.
Jan. 21 '19 Mohamed El Shatshat 1 min
The Neopixel LED ring came in and is quite bright. It also fits over the camera nicely, allowing us to change our mounting strategy.
Jan. 18 '19 Herman Grewal 5 min
The material for the enclosure design is now sourced from the Engineering Machine Shop. Special thanks to