Create AI software to be used with a CAMERA
Budget: $750 – $1,500 USD
I am looking for a skilled developer to create AI software to be used with a camera. The software should have the following functionality:
- Image classification
- The ability to recognize 1-5 different objects or faces
- High accuracy in its recognition capabilities
Ideal Skills and Experience:
- Experience in developing AI software
- Proficiency in image classification algorithms
- Familiarity with camera integration
- Strong understanding of machine learning and computer vision techniques
The problem we currently face: We go to a local bank, purchase $1000 worth of US pennies, then go home and hand search each and every coin, looking for key dates. (IF WE FIND A SUSPECTED KEY DATE, WE THEN HAVE TO TAKE EVEN MORE TIME, AND MATCH THAT KEY DATE TO OUR LIST TO SEE IF IT IS IN FACT WORTH REMOVING FROM THE PILE).
It's time consuming, difficult, and messy.
Proposed solution: We would like to take the same $1000 worth of pennies, and place them out in front of a camera, and then have the AI software (which we need generated), to identify those pieces and match it against the preloaded list we will have ability to load to the newly created AI software.
--
bonus feature: It would be nice to have the ability to turn all coins "right side up". I think we could achieve this by the following method, but the ai would have to allow for this feature:
Solution: The software would turn on "a switch" that we have hooked up to a small small vacuum pump with nozzel. As the coins move by on a conveyor belt in single file fashion, if the AI and CAMERA detect a coin is "face down", the pump would quickly activate the vacuum, which would create a suction on the coin, lift it a few mm off the belt, where another lever would "catch" the edge of the coin, (flipping it) "right side up", back down on the conveyor belt.
This way, we wouldn't have to go through and flip all the "face down" coins manually.
--
Additional Feature: Once all coins are "right side up", they would continue on their "conveyor belt journey", where all key dates would be identified by the AI/CAMERA. If a key date is located (based on the software we preloaded), it would activate another switch which we would have hooked up to another vacuum robot, which would "remove this coin" from the operation feed.
This database will need to have the ability to load photos of each coin too. (The idea for the ability to add photos, is for the software to identify any "errors". IE; mint mark errors, or any other flawed designs from the mint.
so a 2000 US penny from the Philadelphia mint, would have a photo loaded into the software. 99% of the coins would likely match this photo.... but if a variation of this photo is detected, the software needs to immediate alert the AI to this anomaly, because it's an "ERROR coin" from the mint.
- Image classification
- The ability to recognize 1-5 different objects or faces
- High accuracy in its recognition capabilities
Ideal Skills and Experience:
- Experience in developing AI software
- Proficiency in image classification algorithms
- Familiarity with camera integration
- Strong understanding of machine learning and computer vision techniques
The problem we currently face: We go to a local bank, purchase $1000 worth of US pennies, then go home and hand search each and every coin, looking for key dates. (IF WE FIND A SUSPECTED KEY DATE, WE THEN HAVE TO TAKE EVEN MORE TIME, AND MATCH THAT KEY DATE TO OUR LIST TO SEE IF IT IS IN FACT WORTH REMOVING FROM THE PILE).
It's time consuming, difficult, and messy.
Proposed solution: We would like to take the same $1000 worth of pennies, and place them out in front of a camera, and then have the AI software (which we need generated), to identify those pieces and match it against the preloaded list we will have ability to load to the newly created AI software.
--
bonus feature: It would be nice to have the ability to turn all coins "right side up". I think we could achieve this by the following method, but the ai would have to allow for this feature:
Solution: The software would turn on "a switch" that we have hooked up to a small small vacuum pump with nozzel. As the coins move by on a conveyor belt in single file fashion, if the AI and CAMERA detect a coin is "face down", the pump would quickly activate the vacuum, which would create a suction on the coin, lift it a few mm off the belt, where another lever would "catch" the edge of the coin, (flipping it) "right side up", back down on the conveyor belt.
This way, we wouldn't have to go through and flip all the "face down" coins manually.
--
Additional Feature: Once all coins are "right side up", they would continue on their "conveyor belt journey", where all key dates would be identified by the AI/CAMERA. If a key date is located (based on the software we preloaded), it would activate another switch which we would have hooked up to another vacuum robot, which would "remove this coin" from the operation feed.
This database will need to have the ability to load photos of each coin too. (The idea for the ability to add photos, is for the software to identify any "errors". IE; mint mark errors, or any other flawed designs from the mint.
so a 2000 US penny from the Philadelphia mint, would have a photo loaded into the software. 99% of the coins would likely match this photo.... but if a variation of this photo is detected, the software needs to immediate alert the AI to this anomaly, because it's an "ERROR coin" from the mint.