[Fix] prob computation for parseq and vitstr models

doctr/models/recognition/parseq/pytorch.py

@@ -393,18 +393,17 @@ def __call__(
     ) -> List[Tuple[str, float]]:
         # compute pred with argmax for attention models
         out_idxs = logits.argmax(-1)
-        # N x L
-        probs = torch.gather(torch.softmax(logits, -1), -1, out_idxs.unsqueeze(-1)).squeeze(-1)
-        # Take the minimum confidence of the sequence
-        probs = probs.min(dim=1).values.detach().cpu()
+        preds_prob = torch.softmax(logits, -1).max(dim=-1)[0]
 
         # Manual decoding
         word_values = [
             "".join(self._embedding[idx] for idx in encoded_seq).split("<eos>")[0]
             for encoded_seq in out_idxs.cpu().numpy()
         ]
+        # compute probabilties for each word up to the EOS token
+        probs = [preds_prob[i, : len(word)].mean().item() for i, word in enumerate(word_values)]
 
-        return list(zip(word_values, probs.numpy().tolist()))
+        return list(zip(word_values, probs))
 
 
 def _parseq(

doctr/models/recognition/parseq/tensorflow.py

@@ -421,10 +421,7 @@ def __call__(
     ) -> List[Tuple[str, float]]:
         # compute pred with argmax for attention models
         out_idxs = tf.math.argmax(logits, axis=2)
-        # N x L
-        probs = tf.gather(tf.nn.softmax(logits, axis=-1), out_idxs, axis=-1, batch_dims=2)
-        # Take the minimum confidence of the sequence
-        probs = tf.math.reduce_min(probs, axis=1)
+        preds_prob = tf.math.reduce_max(tf.nn.softmax(logits, axis=-1), axis=-1)
 
         # decode raw output of the model with tf_label_to_idx
         out_idxs = tf.cast(out_idxs, dtype="int32")
@@ -434,7 +431,10 @@ def __call__(
         decoded_strings_pred = tf.sparse.to_dense(decoded_strings_pred.to_sparse(), default_value="not valid")[:, 0]
         word_values = [word.decode() for word in decoded_strings_pred.numpy().tolist()]
 
-        return list(zip(word_values, probs.numpy().tolist()))
+        # compute probabilties for each word up to the EOS token
+        probs = [preds_prob[i, : len(word)].numpy().mean().item() for i, word in enumerate(word_values)]
+
+        return list(zip(word_values, probs))
 
 
 def _parseq(

doctr/models/recognition/vitstr/pytorch.py

@@ -159,18 +159,17 @@ def __call__(
     ) -> List[Tuple[str, float]]:
         # compute pred with argmax for attention models
         out_idxs = logits.argmax(-1)
-        # N x L
-        probs = torch.gather(torch.softmax(logits, -1), -1, out_idxs.unsqueeze(-1)).squeeze(-1)
-        # Take the minimum confidence of the sequence
-        probs = probs.min(dim=1).values.detach().cpu()
+        preds_prob = torch.softmax(logits, -1).max(dim=-1)[0]
 
         # Manual decoding
         word_values = [
             "".join(self._embedding[idx] for idx in encoded_seq).split("<eos>")[0]
             for encoded_seq in out_idxs.cpu().numpy()
         ]
+        # compute probabilties for each word up to the EOS token
+        probs = [preds_prob[i, : len(word)].mean().item() for i, word in enumerate(word_values)]
 
-        return list(zip(word_values, probs.numpy().tolist()))
+        return list(zip(word_values, probs))
 
 
 def _vitstr(

doctr/models/recognition/vitstr/tensorflow.py

@@ -164,10 +164,7 @@ def __call__(
     ) -> List[Tuple[str, float]]:
         # compute pred with argmax for attention models
         out_idxs = tf.math.argmax(logits, axis=2)
-        # N x L
-        probs = tf.gather(tf.nn.softmax(logits, axis=-1), out_idxs, axis=-1, batch_dims=2)
-        # Take the minimum confidence of the sequence
-        probs = tf.math.reduce_min(probs, axis=1)
+        preds_prob = tf.math.reduce_max(tf.nn.softmax(logits, axis=-1), axis=-1)
 
         # decode raw output of the model with tf_label_to_idx
         out_idxs = tf.cast(out_idxs, dtype="int32")
@@ -177,7 +174,10 @@ def __call__(
         decoded_strings_pred = tf.sparse.to_dense(decoded_strings_pred.to_sparse(), default_value="not valid")[:, 0]
         word_values = [word.decode() for word in decoded_strings_pred.numpy().tolist()]
 
-        return list(zip(word_values, probs.numpy().tolist()))
+        # compute probabilties for each word up to the EOS token
+        probs = [preds_prob[i, : len(word)].numpy().mean().item() for i, word in enumerate(word_values)]
+
+        return list(zip(word_values, probs))
 
 
 def _vitstr(