[FFmpeg-devel] [PATCH 3/4] avfilter/dnn_backend_torch: add support for dynamic input shapes

[Raja Rathour via ffmpeg-devel <ffmpeg-devel@ffmpeg.org>]

---
 libavfilter/dnn/dnn_backend_torch.cpp | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/libavfilter/dnn/dnn_backend_torch.cpp b/libavfilter/dnn/dnn_backend_torch.cpp
index fee1cfb3a7..4f7ae17aab 100644
--- a/libavfilter/dnn/dnn_backend_torch.cpp
+++ b/libavfilter/dnn/dnn_backend_torch.cpp
@@ -206,8 +206,12 @@ static int fill_model_input_th(THModel *th_model, THRequestItem *request)
     size_t cur_size = input.dims[height_idx] * input.dims[width_idx] *
                       input.dims[channel_idx] * sizeof(float);
 
-    // Initial allocation
-    if (!infer_request->input_data) {
+    /**
+     * Dynamic Resizing Logic:
+     * Only reallocate if the existing buffer is too small or doesn't exist.
+     */
+    if (!infer_request->input_data || infer_request->input_data_size < cur_size) {
+        av_freep(&infer_request->input_data);
         infer_request->input_data = av_malloc(cur_size);
         if (!infer_request->input_data)
             return AVERROR(ENOMEM);
-- 
2.51.0

_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-leave@ffmpeg.org