Administrator/opencv-MIRROR
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-11-09 18:04:29 -06:00
Code Issues Packages Projects Releases Wiki Activity

dnn(torch): validate nn.Padding dimension and pad bounds

Browse Source 
Validate the one-based Padding dimension before converting it to an
array index, and reject the minimum integer pad before negating it.
Add a serialized Torch7 regression test for valid and invalid bounds.

Signed-off-by: Nathan Keys <professor.moody@pm.me>
This commit is contained in:
Nathan Keys
2026-10-28 22:02:46 -05:00
parent fa3037533e
commit 50cfa93293
2 changed files with 81 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
modules/dnn/src/torch/torch_importer.cpp
View File

@@ -783,8 +783,12 @@ struct TorchImporter
if (scalarParams.has("nInputDim"))
layerParams.set("input_dims", scalarParams.get<int>("nInputDim"));
int dim = scalarParams.get<int>("dim") - 1; // In Lua we start from 1.
int rawDim = scalarParams.get<int>("dim");
CV_CheckGT(rawDim, 0, "DNN/Torch: Padding dim must be positive");
CV_CheckLE(rawDim, CV_MAX_DIM, "DNN/Torch: Padding dim exceeds CV_MAX_DIM");
int dim = rawDim - 1; // In Lua we start from 1.
int pad = scalarParams.get<int>("pad");
CV_CheckGT(pad, INT_MIN, "DNN/Torch: Padding pad magnitude overflows int");
AutoBuffer<int> paddingsBuf((dim + 1) * 2);
int *paddings = paddingsBuf.data();

76
modules/dnn/test/test_torch_importer.cpp
View File

@@ -302,6 +302,82 @@ TEST_P(Test_Torch_layers, net_padding)
runTorchNet("net_spatial_reflection_padding", "", false, true);
}
TEST(Test_Torch, padding_rejects_invalid_dim_and_min_pad)
{
auto appendInt32 = [](std::string& out, int32_t v) {
out.append(reinterpret_cast<const char*>(&v), sizeof(v));
};
auto appendFloat64 = [](std::string& out, double v) {
out.append(reinterpret_cast<const char*>(&v), sizeof(v));
};
auto appendString = [&](std::string& out, const std::string& s) {
appendInt32(out, static_cast<int32_t>(s.size()));
out.append(s);
};
auto makePaddingT7 = [&](int32_t dimVal, int32_t padVal, int32_t nInputDim) {
std::string buf;
appendInt32(buf, 4); // TYPE_TORCH
appendInt32(buf, 1); // object index
appendString(buf, "V 1");
appendString(buf, "nn.Padding");
appendInt32(buf, 3); // TYPE_TABLE
appendInt32(buf, 2); // table index
appendInt32(buf, 4); // 4 pairs
appendInt32(buf, 2); // TYPE_STRING
appendString(buf, "dim");
appendInt32(buf, 1); // TYPE_NUMBER
appendFloat64(buf, static_cast<double>(dimVal));
appendInt32(buf, 2); // TYPE_STRING
appendString(buf, "pad");
appendInt32(buf, 1); // TYPE_NUMBER
appendFloat64(buf, static_cast<double>(padVal));
appendInt32(buf, 2); // TYPE_STRING
appendString(buf, "nInputDim");
appendInt32(buf, 1); // TYPE_NUMBER
appendFloat64(buf, static_cast<double>(nInputDim));
appendInt32(buf, 2); // TYPE_STRING
appendString(buf, "value");
appendInt32(buf, 1); // TYPE_NUMBER
appendFloat64(buf, 0.0);
return buf;
};
auto loadPaddingBytes = [](const std::string& bytes) {
const std::string path = cv::tempfile(".t7");
FILE* fp = fopen(path.c_str(), "wb");
CV_Assert(fp != NULL);
const size_t written = fwrite(bytes.data(), 1, bytes.size(), fp);
fclose(fp);
CV_Assert(written == bytes.size());
try
{
Net net = readNetFromTorch(path);
std::remove(path.c_str());
return net;
}
catch (...)
{
std::remove(path.c_str());
throw;
}
};
Net validNet;
ASSERT_NO_THROW(validNet = loadPaddingBytes(makePaddingT7(2, 1, 2)));
validNet.setInput(Mat(1, 3, CV_32F, Scalar(2.0f)));
Mat out = validNet.forward();
ASSERT_EQ(out.dims, 2);
EXPECT_EQ(out.rows, 1);
EXPECT_EQ(out.cols, 4);
EXPECT_FLOAT_EQ(out.at<float>(0, 0), 2.0f);
EXPECT_FLOAT_EQ(out.at<float>(0, 3), 0.0f);
EXPECT_THROW(loadPaddingBytes(makePaddingT7(0, 1, 2)), cv::Exception);
EXPECT_THROW(loadPaddingBytes(makePaddingT7(-1, -1, 2)), cv::Exception);
EXPECT_THROW(loadPaddingBytes(makePaddingT7(INT_MIN, 1, 2)), cv::Exception);
EXPECT_THROW(loadPaddingBytes(makePaddingT7(CV_MAX_DIM + 1, 1, 2)), cv::Exception);
EXPECT_THROW(loadPaddingBytes(makePaddingT7(2, INT_MIN, 2)), cv::Exception);
}
TEST_P(Test_Torch_layers, net_non_spatial)
{
#if defined(INF_ENGINE_RELEASE) && ( \