Deep learning algorithms are used to detect as well as create deepfake images and videos. Images and videos are often used as evidence in police investigations and courts to resolve legal cases since they are considered to be reliable sources. However, deepfake technology increases the development of fake videos, and this may lead to image or video evidence unreliable. This paper aims to qualitatively compare deep learning algorithms and frameworks. To detect real and fake images or videos, various detection algorithms have been proposed after deepfakes were introduced. The current deepfake detection algorithms detect the deepfakes by eye blinking, eye teach and facial texture, head poses, face warping artifacts, eye color, lip movements, audio speakers, reflections in the teeth, spatiotemporal features, and capsule forensics. Deepfake detection algorithms and deep learning frameworks are selected and compared. Deep learning frameworks with different performance and features such as TensorFlow, CNTK, Caffe, Torch, Chainer, and Theano are compared. This helps to use appropriate deep learning algorithms and frameworks for deepfake detection.