Optical film provides anti-glare, anti-reflective, and protective features for cell phones and electronic displays such as LCD screens. Due to increased use of optical film, it is challenging for manufacturers to increase their efficiency in producing better quality films. Even a micro scratch in a high-end application of film can lead to a total failure of the display. Optical and visual methods are typically employed to detect defects, but these methods have limitations such as viewing angle artifacts and design of proper illumination source. This paper describes research to utilize artificial neural networks as a non-destructive defect detection model for predicting the presence of pinhole defects in film. An infrared camera captured the thermal response of optical film subjected to heating and cooling. Pinhole defects of various sizes (0.03mm, 0.08mm, 0.2mm, 0.4mm, 0.7mm, 1mm, 2mm, 3mm, 4mm) were investigated. Pinhole defects are one of the most common types of optical film defects. For the process of identification, thermal differences of defective and defect-free regions were investigated. An Artificial Neural Network was trained to use average absolute temperature difference and cooling rate to predict the presence of a defect. The ANN model was trained and verified using separate data sets. The ANN model was able to classify defective and non-defective samples with a 77.8% accuracy rate. The regression coefficient was 0.5874. These results suggest that artificial neural networks can be used for detecting pinhole defects.