Comparison of Options

Model Architecture Size (Approx.) Inference Speed Use Case
osnet_x0_25_market1501.pt OSNet (tiny) ~4 MB ⚡ Fastest Best for CPU/low-resource
lmbn_n_cuhk03_d.pt LMBN (lightweight) ~10 MB Fast Good alternative to OSNet
mobilenetv2_x1_4_msmt17.engine MobileNetV2 ~15 MB ⚡⚡ Fast (GPU) GPU-optimized (TensorRT)
resnet50_msmt17.onnx ResNet50 ~90 MB Slow Heavyweight, avoid for edge
clip_market1501.pt CLIP-based ~300 MB+ Very Slow Heavy, avoid unless required