Improve card classification

README.md

@@ -15,7 +15,7 @@ ResNet18 ONNX model that classifies a cropped image of a SET® card into its fou
 
 ## Input / output
 
-- Input: RGB card crop, resized to 224x320, normalized with ImageNet mean/std (`mean=[0.485, 0.456, 0.406]`, `std=[0.229, 0.224, 0.225]`).
+- Input: RGB card crop, resized to 112x160, normalized with ImageNet mean/std (`mean=[0.485, 0.456, 0.406]`, `std=[0.229, 0.224, 0.225]`).
 - Output: four classification heads, in order count/color/fill/shape, each one of its three SET values plus `other` (low-confidence / non-card crop).
 - `onnx/model.onnx` (fp32) and, if present, `onnx/model_quantized.onnx` (int8).
 

onnx/model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:84c07dd9f985855bf0e58db34e71dba27a0f7706c2b19f346b0e4cd105ae4c04
-size 115518
+oid sha256:a7ce4de08d94db7be5ff9cd1df8e0680a874ea6df0c2c38b45c95273c319a277
+size 118518

onnx/model.onnx.data

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:39ad86e43c1c7f2e10603f51f9c2cb0c5fbb51e20e6f8f0c5b0b313ae2f28bef
+oid sha256:94b7b00270c2c6cd2883cb1bee18afd9292c2aea6d3d8f51ef76a46d42d7cad9
 size 44761088

onnx/model_quantized.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c11b615d7e86b04d393273b75d314ec9c8bc3f9380e972be7fcba1abdb22a6fd
-size 11370938
+oid sha256:984f690de4f073f81a7d9dec46e175a3f83d8e0b227c9672ab7d418dce331c51
+size 11372823