Deep learning based strategies have shown exceptionally huge advancements in accuracy and quick decision making for applications in intelligent or autonomous driving vehicles. Pedestrian recognition, having applications in autonomous or intelligent vehicles, is one of the vital applications of object detection, an area of basic ongoing research in computer vision. From the last couple of decades, pedestrian recognition has played a vital role in numerous real time applications such as collision avoidance systems for smart or intelligent vehicles, intelligent observation cameras, and domestic security frameworks. The proposed methodology of this work suggests utilizing a dash camera to support a model that identifies all humans that might come in the way of a moving autonomous or driven vehicle from images captured with the dash camera. This algorithm proposed model is based on TensorFlow Human Detection API and is compared to the Histograms of Oriented Gradients (HOG) for human detection. Based on the research, the accuracy and efficiency of the proposed model to detect human was much better than to the other models. The defined method in this paper had an accuracy of around 98% as compared to 84% for the Histograms of Oriented Gradients detection model. The proposed model was faster on average as compared to the Histograms of Oriented Gradients detection model for any given picture on the same framework for training to testing because the proposed model took 11 seconds on average for processing one picture whereas the HOG method took more than 40 seconds on average. The proposed model gives “boxes” as outputs around the humans detected in the images and the result of this model is that it is accurately able to recognize pedestrians with the high accuracy of 98%. The samples on which the algorithm was tested are low-quality pictures taken from low-cost cameras, needing exceptionally less computing power, obviating need for expensive components. This low-cost proposed system will thus permit a dash-cam based ...