There are three ways to get a vision capability: buy a standard product, have a system custom built, or develop it with your own staff. Each is right in some circumstances. Choosing well comes down to how unusual your problem is and how central it is to your business.
When a standard product is the right answer
- The task is common: reading codes, checking presence, measuring simple dimensions, inspecting standard circuit boards.
- A product exists that was designed for it.
- Your part and your environment are within what that product expects.
- You need it working in weeks.
If a standard product does the job, buy it. Custom development of something that is already on the shelf is rarely justified.
Signs that you need something custom
- You have tried standard products and they miss the defect or reject too many good parts.
- The part is unusual in size, material, surface or geometry.
- The measurement needs more accuracy than general-purpose sensors provide.
- The environment is difficult: underwater, outdoors, in motion, at night, at high temperature.
- Several sensors must be combined, for example visible and infrared, or camera and sonar.
- It must run in real time on a small, low-power embedded platform.
- The vision capability is part of a product you sell, and you need to own it.
Comparing the three routes
| Standard product | Custom, by a specialist | In-house development | |
|---|---|---|---|
| Time to first result | Weeks | Months | Longest, if the team is new to vision |
| Initial cost | Lowest | Higher, one-time | Salaries, tools and time |
| Fit to the problem | As far as configuration allows | Designed for it | Designed for it |
| Technical risk | Low if it fits | Managed by staged feasibility | Depends on team experience |
| Ownership of the result | Vendor’s product | By agreement, often yours | Yours |
| Ongoing support | Vendor | Developer, or transferred to you | Your team |
What a custom project involves
- Requirements: what must be detected or measured, how accurately, how fast and under what conditions.
- Feasibility: a short study using real samples or data, to establish whether the approach works and what performance is realistic.
- Prototype: working hardware and software that can be tested in the real environment.
- Product: design for reliability, cost and manufacture; documentation; validation.
- Transfer and support: training your people, handing over what was agreed, and supporting it afterwards.
Each stage ends with a decision. A project structured this way limits what you spend before you know the answer to the main technical question.
Questions to settle early
- Who owns the design, the software and any inventions?
- What are the acceptance criteria, in measurable terms, and how will they be tested?
- What data or samples can you provide, and when?
- Who maintains the system in five years?
- Is the developer able to build hardware as well as write algorithms?
Choosing a development partner
Algorithms that work in a laboratory and products that work in the field are separated by a great deal of engineering. Look for a partner who has shipped products, not only prototypes; who can handle optics, electronics and mechanics as well as software; and who will begin with a feasibility study instead of a large fixed quotation.
How VisionMaster works
We design and manufacture our own inspection systems, and we apply the same optics, 3D sensing and Vision AI engineering to customer projects, from a first feasibility study through to a delivered product.