First, give conclusions that can be used for decision-making
The data should be prepared in such a way as to identify the target, the action and the consequences of the error, and then design the collection and labelling. Classification, detection, partitioning and OCR need different forms of labelling, and serious and general deficiencies should be separated.
What conditions need to be identified before judgement is made?
The same question may have different answers under different business, data and project phases. It is suggested that the following conditions be checked and that the common findings on the web be incorporated into their own projects.
Suggested order of advance
First, we'll be clear about the target and the border.
(c) Identifying objects, categories, risks, final actions and unjudgeable rules.
Validation Key Dependence
A representative sample is taken and a data check is performed by the field conditions and by batches.
Development of assessable outcomes
A small double-marking exercise is conducted to check consistency before expansion.
Make sure you decide the next step with the real results.
Complement the exception by model type and keep independent final test set.
How do you understand it in the actual business?
The plant collects 100,000 images of normal products, but only dozens of real defects, all of which come from the same camera. The data are not sufficient to ensure the effectiveness of production. The team should first redefine the level of defects and supplement the unusual samples of different production lines, light and batches, and establish reliable small-scale assessments, if necessary, through on-site test mining and expert review.
The easiest pit to step on.
Only the total number of pictures is sought, and no source or conditions are recorded.
The labeled personnel were trained directly on the inconsistent definition of defects
Put the same video in the same frame and both in the training and testing.
How should we end up receiving and confirming?
The data phase should contain a description of the type of delivery, the scope of collection, the standard description, sample distribution, quality sampling, authorized boundaries and version records.
When preparing to communicate with suppliers or internal teams, it is recommended that current processes, representative samples, existing systems, planning time and budget levels be brought. First, the unknown items are clearly marked, and then the decision is made to use diagnostics, PoC, fixed-range projects or ongoing research and development, which is usually more reliable than a direct demand for a price and duration without borders.