The primary methods to mitigate wildlife-vehicle conflict are to provide individual animals with safe passages across roadways and to warn drivers and animals of impending crashes. The second method has had one successful implementation in the US, the thermal sensor based “wildlife crosswalk” across SR 260 in Arizona. As wildlife move toward or onto roads, they pose an immediate and potentially serious collision threat to drivers. The roadside method to mitigating collision risks between vehicles and animals involves detecting the animals as they approach or enter the transportation route and classifying the risk based on trajectory, animal size, and animal speed. We will describe two methods for detecting and classifying wildlife in the roadside: 1) conventional camera traps combined with a customized artificial-intelligence system for immediate classification of wildlife; and 2) a rapid and accurate method to detect and classify wildlife using a combination of CCTV (or still cameras) and AI support. 1) The camera trap method depends on a cell- (HCO Wireless) or wifi- (Buckeye Cams) communicating camera trap to capture and transmit the image to a local or remote server holding the AI system. We used 2 AI methods to detect and classify wildlife – a) a custom convolutional neural network (CNN) built on a Keras API (Google) that could accurately classify deer in an image; and b) the MegaDetector (Microsoft) system, preceded by image processing to enhance the animal in an image. The custom CNN tool for deer was very fast (milliseconds) and highly-accurate (~95%) if trained to the camera view. The MegaDetector tool was much slower (2 seconds), but very accurate in determining if an animal was in the view (~99%), but not the type of animal. 2)The CCTV method is being deployed for Nevada Department of Transportation as part of a multi-sensor station (including radar, LiDAR, and thermal sensors) along USA Parkway in North-Central Nevada. We combined a commonly-available CCTV (SCW Inc.) with the commercially-available AI from the same company, as well as a stream to a custom tool using YOLO v3 (https://pjreddie.com/darknet/yolo/). The YOLO v3 has been proven to be fast and accurate and is most appropriate for streaming imagery, for example from CCTV. The price point of the still camera system (<$1,000/unit) and large-animal detection range (100 m daytime, 30 m night) equates to a day-time effectiveness of ~$10/m protected roadside, or close to $17,000/mile on one side of the road. The price point of the CCTV system ($2,000/unit hardware, $300/year commercial service) and large-animal detection range (>150 m day and night) equates to an effectiveness of ~$12-15/m protected roadside, or up to $26,000/mile on one side of the road. Either option is considerably less expensive than fencing, though night-time effectiveness may depend on the actual unit deployed.