Consistent compliance with quality requirements of the product becomes the key to the sustainable competitiveness for food producers competing in domestic and export markets. More and more markets become so quality sensitive that even a small number of defective product packages might lead to recalls of the whole batch and damaged relationship with clients. It is especially true for Japanese and South Korea markets. While the source of non-compliance might be any production process, including all upstream processes before packaging, improved quality inspection systems at the end of the production line to eliminate non-compliant packaging becomes extremely important. In this paper, a machine vision system was developed for quality inspection of the packaging to visually identify contamination and quality of sealing and label by the smart camera to detect the visual non-compliance. The vision system consists of an inspection camera for capturing the image of thepackage and an image analysis software with Machine Learning capabilities to identify misplaced, missing and damaged labels orsealing problems. The inspection system achieved necessary functionality and precision - only 0.5% of contaminated end sealing packages with particles longer than 1 mm were allowed, speed of inspection - up to 100 packages per minute, time to reject - the time from bypassing the scanner to the rejection - max. 400 milliseconds. Further development of system accuracy and speed using Machine Learning reached 120pcs/minute and continues to increase while requiring less time and labour than existing inspection methods.