Works with yolov4-tiny and TRT7/8

[Nuzhny007]

1. Merged TRT8 branch
2. Fixed some errors for TRT7
3. Added yolov4-tiny and fixed it work
4. Some speedup for blob creation
5. Some code cleanup

[JoeyZhu]

yolov4-tiny runtime failed with

et3DTensorVolume[1]: 43095 != 33150

test envirment: cuda 11.4.0, trt 8.2.5.1 test with cfg and weight from: https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-tiny.cfg https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.weights

Detailed log: joey@j-red:~/samplecodes/2022/yolo-tensorrt_o/build$ ./yolo-trt Loading TRT Engine... WARNING: TensorRT was linked against cuBLAS/cuBLASLt 11.6.5 but loaded cuBLAS/cuBLASLt 11.5.2 Loading Complete! WARNING: TensorRT was linked against cuBLAS/cuBLASLt 11.6.5 but loaded cuBLAS/cuBLASLt 11.5.2 get3DTensorVolume[1]: 43095 != 33150 yolo-trt: /home/joey/samplecodes/2022/yolo-tensorrt_o/modules/yolo.cpp:1268: bool Yolo::verifyYoloEngine(): Assertion volSize == tensor.volume && "Tensor volumes dont match between cfg and engine file \n"' failed. Aborted (core dumped) `

yolov4 and yolov5 test ok

[Nuzhny007]

But with my fork ( https://github.com/Nuzhny007/yolo-tensorrt ) it works correctly? I think that this bug I fixed.

[JoeyZhu]

But with my fork ( https://github.com/Nuzhny007/yolo-tensorrt ) it works correctly? I think that this bug I fixed.

yes, it worked, great work!

but two small compiler error needs to be fixed before its working :smile: