Starting with DetectorPack 3.10, a utility was added to the RR LPR add-on (see Installing DetectorPack add-ons), which allows you to create GPU neural network caches without using Axxon One. The presence of the cache speeds up the initialization and optimizes video memory consumption.

To optimize the operation of the Vehicle recognition RR detector on GPU, do the following:

1. Stop the server (see Stopping the server).
   

   If the system has the software running on GPU, it is necessary to stop its operation.

2. Create the GPU_CACHE_DIR system variable (see Appendix 9. Creating system variable) by specifying the path to the cache location with an arbitrary folder name in the Variable value field. For example, D:\AO_GPU_cache. The specified directory will store the cache for all used detectors and neural networks.

3. To call the utility, open in the command line: C:\Program Files\Common Files\AxxonSoft\DetectorPack\RRGpuCacheGenerator.exe and press Enter.

4. Select the required cache sets by specifying the y or n value:

   1. lpr recognition—recognition of a vehicle license plate;
      

      Vehicle license plate recognition isn't available for the Vehicle recognition RR detector. Specify the n value for this item.

   2. vehicle recognition—recognition of a vehicle;
      
   3. When you select the vehicle recognition cache set without lpr recognition, you must choose the territory where vehicle recognition is going to be performed:
      
      1. CIS countries—1;
      2. Other countries—2.
         
5. Select one of the accuracy and performance values by specifying the corresponding number:
   
   1. Medium accuracy, high performance—1, provides medium recognition accuracy and high performance of the GPU (less load on the GPU);
   2. High accuracy, medium performance—2, provides high recognition accuracy and medium performance of the GPU (high load on the GPU).
      
6. Specify the ID of the required GPU (see Selecting Nvidia GPU when configuring detectors).
7. Press Enter.

As a result, the process of creating cache starts, which takes about 30 minutes. The duration of the process depends on the selected types of cache, number of countries, and Nvidia GPU model.

Optimizing the operation of the Vehicle recognition RR detector on GPU is complete.